# OrgSync help The text copy of https://orgsync.co/help, rebuilt with every release. It explains how OrgSync works: setting up, connecting AI apps, each command, sharing and approval, plans and usage, troubleshooting and what's new. It describes the product, not your company: your company's facts come from OrgSync's tools in your AI app. ## Getting started as an owner 1. **Sign up.** Go to [Sign in](https://orgsync.co/sign-in) and choose **Continue with Google**, or create an account with your email and a password. OrgSync emails you a link to verify the address first. 2. **Create your company.** Three short steps: its name, where it's based and its stage, and, if you like, where you want help first. You can describe what it does now or later; your team's AI uses that description as background. Change any of it in **Settings › Company**. 3. **Start your free trial or choose a plan.** The setup guide opens next and asks this first. Until setup is done, it opens again each time you come back, unless you choose **Later** on its Home card; **Open setup** on Home brings it back any time. The free trial takes a card but charges nothing until it ends, and you can cancel before then. You can also decide later in **Settings › Billing**. See [Plans, free trial and usage](https://orgsync.co/help#pricing). 4. **Describe your role.** Optional, and takes a minute. Fill in the fields, paste a job description, or describe it in a short chat that drafts your role for you to check and edit. Nothing is saved until you save or confirm. It helps your AI focus on your work. 5. **Connect your AI.** Pick your AI app and follow its steps, then run `/orgsync:get-started` in a new chat. It checks the connection, welcomes you by role and asks three quick choices: what OrgSync should help with first, whether your AI offers to share decisions or waits for you to ask, and whether it gives you a short brief when you start working. The guide shows when it works. See [Connect your AI](https://orgsync.co/help#connect). 6. **Bring your company in.** In your AI app, run `/orgsync:import` to bring in decisions from tools your AI app already reaches, from your past chats where your AI app can search them, or by answering a few questions. You check every draft, then one approval shares the ones you keep. Home shows how much your AI can draw on across eight items: sales and marketing, finance, legal, product, hiring and people, operations, a company goal and everyone's role. The step is done at 6 of 8, so areas that don't apply yet can wait. See [Bring in existing material](https://orgsync.co/help#import) and [Share what you know](https://orgsync.co/help#interview). 7. **See your picture and ask something real.** Run `/orgsync:today` to see what needs you and the decisions that shape your work. Make one update a company decision: say “Make that a company decision.” and confirm through [OrgSync's prompt](https://orgsync.co/help#sharing) or [chat confirmation](https://orgsync.co/help#chat-confirmation). In a new chat, ask something that depends on it: your AI answers with a citation. 8. **Invite teammates.** In **Settings › Team**, choose **Invite teammate**. Enter their email and the access they get, and optionally describe their role: its title, what it's for, what they own, and what they decide alone and what they ask about. OrgSync makes an invite link that you send them yourself. It works only for that email, expires in 7 days and is shown only once; choose **Get new link** if it's lost. There's no per-person charge. 9. **Run your weekly review.** Once a week, run `/orgsync:review` to check in on your goals and share the decisions you choose. As an owner, you also see your members' latest reviews. See [Your weekly review](https://orgsync.co/help#review). ### Access - **Owner**: Also manages the team, billing and company settings. - **Decision approver**: Also decides what becomes an official company decision. - **Member**: Uses your company's shared knowledge and shares updates. Access is separate from a person's role, which describes their work and helps their AI focus on it. ## Joining as a member 1. **Open your invitation link.** The owner sends it to you. It works only for the email it was sent to, and expires after 7 days. 2. **Sign in with that email.** Choose **Continue with Google**, or sign in with an existing verified account. Signed in with another account? Choose **Use another account**. 3. **Join.** Check the company, your access and the role the owner described for you, then choose Join. 4. **Connect your AI.** The setup guide opens after you join, and each time you come back until it's done. It shows your role, which only an owner can change, then the steps for your AI app and `/orgsync:get-started`, which checks the connection and shows you around. See [Connect your AI](https://orgsync.co/help#connect). Then ask your AI about something your company has shared. 5. **Your first weeks.** For 45 days after you join, **Your first weeks** on Home lists five things: your role, who you report to, connecting your AI app, your first goal and your first weekly review. Each is ticked only when OrgSync's records show it, such as a saved goal. Your manager and owners see how many are done, never what you read. Choose **Hide from Home** to put it away; it changes nothing you can do. If the link expired, was cancelled or was already used, ask the owner for a new one. ## Connect your AI OrgSync works with these AI apps today. A connection works as you, in one company, and reads only the company knowledge you can see. In OrgSync, **Settings › AI apps** shows these steps with your company's values filled in, each with a Copy button. Here, ``, `` and `` mark where they go. ### Claude Desktop or Claude on the web Desktop app or claude.ai 1. **Add a custom connector.** In Claude Desktop or at claude.ai, open **Customize** › **Connectors** › **Add custom connector**. Name it OrgSync and paste this server URL: `` 2. **Use OrgSync's client ID.** Choose **Sign in now**, **Use your own OAuth client** and **Streamable HTTP**. Paste this client ID and leave the client secret empty: `` 3. **Sign in.** Add the connector. Your browser opens OrgSync: check that it shows your own name and your company, then allow. If it shows someone else, choose **Sign out** there and sign in as yourself. 4. **Add the OrgSync plugin.** Open **Customize** › **Plugins** › **Add** › **Add marketplace** › **Add from a repository**, paste this, choose **Sync**, then **Add** beside OrgSync. It uses the connector you just added. `https://github.com/orgsync-ai/plugins` The plugin teaches Claude how to use OrgSync and adds its commands, such as `/orgsync:get-started`. If the OrgSync plugin shows **Disabled** under **Plugins**, open it and turn it on, then start a new chat. If Claude already has the OrgSync plugin from before, open **Customize** › **Plugins** › **Add** › **Manage marketplaces** › **Check for updates** so the newest commands, such as `/orgsync:get-started`, appear. 5. **Check it.** In a new chat, type `/orgsync:get-started`. On a Claude Team or Enterprise plan, a workspace owner adds the connector before members can connect it. A free Claude plan has one custom connector. ### ChatGPT Desktop app or chatgpt.com 1. **Add a custom MCP server.** In ChatGPT, open **Plugins** › **Add** › **Add custom MCP server**. Name it OrgSync connection, choose **Server URL** and paste this, and keep **Authentication** on OAuth: `` 2. **Use OrgSync's client ID.** Open **Advanced OAuth settings**. Keep **User-Defined OAuth Client**, paste this OAuth client ID and leave the client secret empty. The callback URL should read `https://chatgpt.com/connector_platform_oauth_redirect`. Under **Scopes**, type `offline_access` in **Base scopes** so ChatGPT stays signed in. `` 3. **Sign in.** Tick **I understand and want to continue**, choose **Create as a plugin**, then **Continue to OrgSync connection**. Your browser opens OrgSync: check that it shows your own name and your company, then allow. If it shows someone else, choose **Sign out** there and sign in as yourself. If ChatGPT says an app with this name already exists, remove the old OrgSync app in ChatGPT's settings first, or give this one a different name. 4. **Add the OrgSync plugin.** Download [orgsync-chatgpt.zip](https://github.com/orgsync-ai/plugins/releases/latest/download/orgsync-chatgpt.zip). Then open **Plugins** › **Add** › **Upload plugin archive**, add it and choose **Install plugin**. It teaches ChatGPT how to use OrgSync. If ChatGPT says it couldn't add the plugin and you added OrgSync before, open the OrgSync plugin under your plugins, choose **More actions** › **Upload new version**, upload the same file, then install it. After tool changes, open the OrgSync connection in ChatGPT's settings and choose **Refresh tools**. 5. **Check it.** In a new chat, type **@OrgSync get started**. If ChatGPT has no **Add custom MCP server**, your ChatGPT plan or workspace doesn't allow custom servers; ask your workspace owner. ### Claude Code Runs in Terminal on macOS 1. **Install the OrgSync plugin.** Run this in Terminal: `claude plugin install orgsync --marketplace orgsync-ai/plugins` It needs Claude Code 2.1.292 or later. The plugin teaches Claude how to use OrgSync and adds its commands, such as `/orgsync:get-started`. 2. **Sign in.** Start Claude Code, type `/mcp`, choose **plugin:orgsync:orgsync**, then **Authenticate**. Your browser opens OrgSync: check that it shows your own name and your company, then allow. If it shows someone else, choose **Sign out** there and sign in as yourself. If you connected OrgSync before, `/mcp` lists that connection instead, and the plugin uses it. 3. **Check it.** Type `/orgsync:get-started`. **Connect without the plugin** 1. **Add OrgSync.** Run this in Terminal: `claude mcp add --transport http --scope user --client-id --callback-port orgsync ` 2. **Sign in.** In Claude Code, type `/mcp`, choose **orgsync**, then **Authenticate**. Your browser opens OrgSync: check that it shows your own name and your company, then allow. If it shows someone else, choose **Sign out** there and sign in as yourself. ### Codex CLI Runs in Terminal on macOS 1. **Add OrgSync.** Run this in Terminal: `codex mcp add orgsync --url --oauth-client-id ` 2. **Sign in.** Run this. Your browser opens OrgSync: check that it shows your own name and your company, then allow. If it shows someone else, choose **Sign out** there and sign in as yourself. `codex mcp login orgsync --scopes offline_access,mcp:status,mcp:submit-draft,mcp:read-context` If Codex started a sign-in by itself in step 1, cancel that one and use this command. 3. **Add the OrgSync plugin.** Run this to teach Codex how to use OrgSync: `codex plugin marketplace add orgsync-ai/plugins && codex plugin add orgsync@orgsync` 4. **Check it.** Start a new Codex session and ask: “Get started with OrgSync.” When the check passes, **Settings › AI apps** shows the connection as **Connected**. Its access lasts a limited time: the connection's page in AI apps shows until when, and then you sign in again. ## Your connection's instructions When your AI app passes on OrgSync's connection instructions, your AI receives your name, your work profile, recent accepted decisions' titles and up to three adopted practice titles. The OrgSync plugin tells it to check company context when a task touches your company's decisions, policies or responsibilities. Start by [connecting your AI and adding the plugin](https://orgsync.co/help#connect); there's no command to turn this on. If your app doesn't pass on the instructions, use `/orgsync:today` to give your AI your picture of the company. Your AI can also start any OrgSync command when you ask in your own words, such as “let's do my weekly review”. These instructions and opening a listed decision cost no company credit. A cited answer uses 1¢. They never give your AI more access than you have or capture your conversation in the background. ## Your picture of the company Use `/orgsync:today` for your picture of the company: what needs you, what changed for you, your focus (up to five active goals), how we work, decisions that shape your work, and your role. It also lists your to-dos that are due or overdue. “How we work” shows adopted practices, with more available when the list is cut short. An older server may leave it unavailable; that is different from having no adopted practices. Your AI writes a short brief from the items it shows. It puts what needs you first and offers to work through it one item at a time: answer or agree to a request, accept a shared update, decide a suggested correction, or share your own draft. Each change still needs your confirmation. It looks back seven days by default. Add a time, such as `/orgsync:today since last Monday`, to choose another starting point, up to 30 days ago. It tells you when a list is cut short. This uses free company reads and no company credit. It never runs a paid cited retrieval or acts on a waiting item without your request. ## To-dos and reminders Ask your AI app to remember something for your work: “remind me Friday at 10 to send Acme the new price”, “every last Friday of the month, check the runway”, or “remind me Thursday unless Sam has submitted the report”. OrgSync keeps it as a to-do that only you can see. Your teammates and your company's Owners cannot read it. A to-do can have a date, a time, a repeat, up to three reminders, and links to a goal, a request, a source or teammates. Times follow your time zone. A repeat follows its schedule or starts again from when you finish it, and a missed one moves to the next date instead of piling up. A title can't hold a link or an email address. **On the website.** On Home, **To-do** at the left edge of the screen (on a phone, in the header) opens your own to-dos for this week, this month or all of them, with overdue ones first. Only you can see this list, and it's read-only: choose **Continue in your AI** to change, reschedule or close one. Your dated to-dos also show in **Calendar**, under **Mine** only, never under **Company** or in **Download calendar**. **Reminders.** At the time you set, the to-do shows as due in `/orgsync:today` and in the Claude Code panel. OrgSync never emails to-dos or reminders; for a phone alert, add it to your phone's reminders too. **Follow-ups that close themselves.** A to-do linked to a request or a goal can wait on it. If Sam submits the report before Thursday, the to-do closes itself instead of coming due. **Handing off and reviewing.** Ask your AI to hand a to-do to a teammate: it becomes a request, which they agree to as usual. Your weekly review shows to-dos that keep slipping and asks whether to drop, defer, hand off or keep them. Done, dropped and handed-off to-dos are deleted 30 days after you close them. To-dos are free and use no company credit. Reminders for life outside work belong in your phone or your own to-do app. ## Ask or assign a teammate Use `/orgsync:request` to ask a teammate for information, a piece of work, a review or a decision, or to act on a request someone sent you, for example `/orgsync:request ask Priya for last week's refund numbers by Friday`. Your AI finds the person, asks only what's missing (what a good answer looks like, by when, who reviews it, which goal it moves) and checks their current load when your company uses Kanban. OrgSync shows you the exact request to confirm before it's created. Everyone in your company can read requests and their replies, so keep confidential work outside OrgSync. Nothing is sent by email or chat: the person sees it at their next OrgSync check-in, so tell them directly if it's urgent. Requests are free. When someone asks you, open it the same way to answer, agree, decline, propose another date, report a blocker, submit your result or review one. Acknowledging an assignment from your manager is separate from agreeing to its terms. See [Weekly contributions with your team](https://orgsync.co/help#collaboration). **Asking your own manager.** Ask your AI to “ask my manager…”: it knows who your manager is from your role. A piece of work needs no goal. A company decision goes to your manager only if they are an Owner or Decision approver; otherwise your AI asks them as a question, and can send the decision to someone who decides. ## See how the team is doing Use `/orgsync:team` to see how your team or company is doing: goals at risk or without a recent check-in, blocked or overdue work, who has or hasn't done a weekly review, and what waits for approval. Add a team, or a person's name to prepare a one-on-one. It uses free reads of records everyone in your company can already see, and shows facts, not scores or rankings of people. It changes nothing; it offers next steps, such as asking someone through a request, only when you ask. ## Turn a meeting into updates Use `/orgsync:meeting` with notes from a meeting or call: paste them, point your AI at a notes page it can already reach, or use the current chat. Your AI prepares three lists: decisions to share, follow-ups as requests to teammates, and goal updates to record in your weekly review. You drop or edit any item, then confirm: one confirmation shares the decisions, and each request has its own. The notes themselves are never shared, and nothing is recorded or read in the background. Each shared decision uses 10¢ of company credit once it's ready for AI. ## Bring in existing material Use `/orgsync:import` in your AI app to bring in company material. You choose a set, such as a folder or channel. Your AI reads it through tools already connected to that conversation and makes short drafts, each with a link to its original. It first shows how much your AI can draw on and what it can reach: tools connected to your AI app, and your past chats where the app can search them. With your yes, it looks through chats from the last 90 days for decisions only, such as prices, policies, owners and commitments, and leaves out personal matters; you see a title and a line for each, never the chat. Where nothing connected holds an area, it asks a few questions instead. Ask it to check originals to compare the sources you shared with where they came from. You see every title and any warning about private material, and can drop or edit drafts before one approval shares a batch of up to 20. Everyone in your company can see what you share. Each draft that becomes ready for AI uses 10¢ of company credit, including during the free trial. OrgSync never connects to those other services itself, imports material you can't read, or shares without your approval. It never updates an import by itself. If a tool is missing, connect it in your AI app's settings, then start again. ## Share what you know Ask `/orgsync:import` to ask you questions, optionally about an area such as pricing or hiring. Your AI asks one question at a time, fitted to your work profile, and turns your answers into private drafts. You can stop at any time. For questions about you, your projects and your goals, run `/orgsync:get-started`: its **About you** step asks them. Each result you want can become a goal that you confirm; OrgSync has no place for projects, so your AI app keeps their details in its own memory if you agree. Check the titles and summaries, drop or edit any, then approve the drafts you keep in one batch. The company reads used to prepare the interview are free. Each shared draft that becomes ready for AI uses 10¢ of company credit. The interview never adds answers you didn't give or shares drafts without your approval. It doesn't make them accepted company decisions; an Owner or Decision approver does that separately. Before plugin 0.19.0 this was `/orgsync:interview`, which is now gone. ## Think through a decision Use `/orgsync:decide` followed by the decision you're considering. Your AI reads relevant decisions, asks OrgSync once for cited context, and lays out what is already decided, who to involve, gaps and options. You make the decision. Opening decisions is free; the cited retrieval uses 1¢ of company credit. If you want to share the outcome, your AI prepares a draft for your approval. A shared update uses 10¢ once ready for AI. It never shares a conversation in the background or changes a decision without approval. If your decision replaces an existing one, it offers a correction to that source. The old decision stays current until an Owner or Decision approver accepts the correction. To share something you already decided in the conversation, ask `/orgsync:decide` to share it; your AI shows the exact text before anything is shared. Before plugin 0.19.0 this was `/orgsync:share`, which leads here for one more version. Ask a question about what your company decided in your own words; `/orgsync:ask` also still works for one more version. If you state something an accepted decision contradicts, your AI says so in one sentence and offers to use the company decision or to flag it as outdated. Flagging reports it under your name and changes nothing until an Owner or Decision approver acts; your choice stands. ## Check the originals Ask `/orgsync:import` to check the originals of sources you shared, or name one to check it. Your AI opens each source and re-reads its original through a tool already connected to your AI app. It checks up to 10 per run and offers the rest afterwards. It reports each as unchanged, changed or unreadable, and shows changed decisions side by side. These reads cost no company credit. If you approve a suggested correction, an Owner or Decision approver still needs to accept it; that publication uses 10¢ once ready. It never syncs or changes sources automatically. A missing original doesn't delete or change the OrgSync source. If a tool is missing, connect it in your AI app's settings; an original with no usable link or path can't be checked. Before plugin 0.19.0 this was `/orgsync:refresh`, which is now gone. ## Choose your next useful goal In your AI app, run `/orgsync:guide` to choose one useful next goal. It starts from your business profile, agreed decisions, current goals and previous review, then asks only for missing details. If company context is missing, it can help you [share what you know](https://orgsync.co/help#interview) or [bring in existing material](https://orgsync.co/help#import). Start with how your company works now. Keep it, tailor it, write your own approach or try an optional example. You can skip adopting a way of working, including when working alone. A label or area alone never makes a practice adopted. It must be a current, ready source accepted as a company decision by an Owner or Decision approver. You confirm every saved goal and shared change in your AI app. To change an existing practice, your AI proposes a correction for separate approval. Later sessions use the current accepted revision. If practices conflict, a person resolves them. The guide uses free company reads. Saving goals and reviews is free; sharing and cited answers use the usual company credit. For finding and winning customers, it can follow [First customers](https://orgsync.co/help#first-customers). Return with [your weekly review](https://orgsync.co/help#review) to compare what you did with what changed. In ChatGPT, type **@OrgSync** before your request. In Codex, ask it to use OrgSync. The guide sends no messages and grants no permission to act outside OrgSync. ## Set and revisit goals Use `/orgsync:goals` to set a goal for your company or for one person, or to revisit one. Each goal has the result you want, how you'll know it was reached and a date. A company goal can have no one responsible; a goal for one person always names who is. Your AI asks for each part and never invents one. You confirm each exact change before it's saved, through your connection's mode: OrgSync's prompt, or a yes in the chat for a chat-confirmation connection. A decline or dismissal changes nothing. Saving a goal is free. When your company has OKR guidance on, a goal can have up to three key results. A measurable one has a unit, a starting value and a target; one with no target is qualitative. Your AI takes the numbers only from you. In your weekly review you record each key result's latest value, and OrgSync works out its progress and the days since the last check-in. Progress describes the goal; it never rates a person. A goal can also carry up to five of your company's tags, as in [Weekly contributions with your team](https://orgsync.co/help#collaboration). Every company member can read goals and their history. Owners and Decision approvers can change any goal and assign active members. Members can change only the person goals they own. Being responsible for a goal gives no permission. A person goal can optionally support one active company goal. You confirm the link or its removal with the exact goal change. If the company goal changes or closes, your AI shows that on the next read or review. Your goal stays as it was until you decide what to do. `/orgsync:today` shows **Your focus**, with up to five active goals. On the website, **Goals** shows them read-only, in sections. Completing or dropping a goal is final. Your AI asks for your judgment of the result, and for your confirmation, before it records either. ## Find your first customers Ask `/orgsync:guide` for help with first customers to plan your next step toward them. Your AI asks what you sell, who your customers are, where they are, and how much time and budget you have. It takes your stage and location from your business profile and asks for anything missing. It then fits one small plan to your stage. A business that is only an idea gets a small plan for talking with customers about one problem. A business that is starting gets one offer, one customer group and one way to reach them. Then it prepares one useful piece, such as interview questions, an offer, outreach wording or a sales-conversation outline. You carry out the outside work yourself. This guide never sends messages, contacts prospects, buys advertising or takes any external action. It gives no legal advice. For questions that depend on your location, such as marketing, privacy, consent, tax or registration rules, it points you to current official guidance or a qualified adviser. If you want to track the plan, your AI proposes one goal with a date and a responsible person. You confirm it as described in [Set and revisit goals](https://orgsync.co/help#goals). Before plugin 0.19.0 this was `/orgsync:first-customers`, which is now gone. ## Your weekly review Run `/orgsync:review` once a week in your AI app. Your AI takes you through these parts, in order: - It shows the priorities from your last review. If you skipped weeks, it calls them that review's priorities, not last week's. - For each goal assigned to you, it asks whether the goal is on track, at risk or off track. It asks separately what you did and what moved the result, because work done doesn't prove a result moved. If the goal has key results, it asks for their latest values, and it points out a goal with no check-in for seven days or more. Owners and Decision approvers can also check in on company goals with no one assigned. - It asks which of this week's decisions you want to share with your company, and prepares a short summary with supporting excerpts for each one you pick. - It asks for a call on each goal: keep going, change approach, stop, or done. - It offers research on your top goal. Search runs only when you ask, using your AI app's own search, and gives at most two sourced ideas. Then it sets one to three priorities for next week; an idea you accept becomes one of them. One confirmation saves your review and shares the decisions you chose. You confirm it through your connection's mode: OrgSync's prompt, or a yes in the chat for a chat-confirmation connection. A decline or dismissal saves nothing, and the decisions stay private. Saving the review is free. Each shared decision uses 10¢ of company credit once it's ready for AI. If some shares fail, your review is still saved, and your AI reports each result. The next review starts from this one's priorities. A stop or done call doesn't close a goal. It needs a separate goal change, which you confirm as described in [Set and revisit goals](https://orgsync.co/help#goals). An Owner also sees members' latest reviews, and who has none yet. Linked check-ins are grouped by the company goal they support, with each member's name and evidence kept separate. Unlinked check-ins remain visible. A changed or closed company goal needs your judgment; it never closes a person's goal automatically. ### Weekly reminder email The reminder is optional and set for each company. In **Settings › Account**, find the section headed “Weekly review reminder for” and your company's name. Tick the box that starts “Email me a weekly reminder”, choose a day, check the time zone (it starts as your browser's; change it if needed, such as America/Toronto), then choose **Save reminder**. OrgSync sends at most one email a week, on your chosen day at or after 9:00 in your time zone. It names the company and contains counts only, never company content. Each email has an unsubscribe link that works in one click, without signing in, and shows a plain confirmation page. You can turn the reminder back on in **Settings › Account**. The reminder only sends an email. It doesn't run the review; you run that yourself. ## Weekly contributions with your team Use `/orgsync:request`, or ask your AI app, to connect a weekly contribution to a goal, name the recipient and reviewer, and agree what evidence will show the result. For example: “Use OrgSync to request customer interview findings from Alex for our retention goal, with the evidence and limitations ready for Friday's review. Show me the exact request before sending.” A peer request becomes your commitment only when you explicitly agree. A direct or active higher reporting manager can assign a contribution immediately. Acknowledging that assignment records receipt. Agreeing to its terms is separate, and a manager-set date stays distinct from a date you agreed. At your next check-in, ask “Use OrgSync to show what needs me, what I am waiting for, blockers and upcoming reviews.” You can propose another date, flag a conflict or blocker, and submit selected evidence. Ask “Read this OrgSync contribution and help me review its latest submission against the agreed criteria.” Accepting a contribution does not accept a company decision or complete a goal. Requests, responses, blockers and reviews are readable by everyone in your company. Prepare privately in your AI app, then confirm what to send. Keep confidential work outside this workflow. An external evidence link grants no new access to its contents. Requests appear at your next successful OrgSync check-in, with active-session updates where supported. A closed AI app does not receive a notification. If a response times out, retry the same confirmed intent or read its receipt before preparing another change. Home has a compact collaboration overview and an expandable agenda, **Calendar** on Home shows request and goal dates, work in progress and decisions taking effect, by month or week, and **Work** shows every request on a board. Detail pages are read-only and provide a prompt you can copy into your AI app. Agenda export downloads a calendar snapshot for the chosen range and audience. Imported copies do not synchronize and cannot be recalled. Owners turn OKR and Kanban guidance on or off in **Settings › Company › Ways of working**, with optional custom instructions for each. Both start enabled. Kanban can also set a work-in-progress limit per person; it never blocks work. Other references, such as Scrum or RICE, change only from an Owner's AI app, and one project reference is on at a time. Guidance creates no goals, assignments or scheduled work, and adopting a company practice still requires human approval. Current company choices and approved practices take priority over packaged instructions. Tags label goals, requests and shared decisions by team, project or topic so people can find work. An Owner's AI app adds and archives them; a Member can ask an Owner through a request. Tags and mentions grant no access, assign no work and record no agreement. ## The Claude Code mod The OrgSync plugin includes a Claude Code mod. After [adding the plugin and connecting OrgSync](https://orgsync.co/help#connect), start a session in the Claude Code CLI or Claude Desktop's Code tab. When facts are new, it shows a notice at session start and gives your AI the same picture. Use `/orgsync:today` with no arguments to open the full picture. When the mod can't draw it, the skill writes it as text. Changes during a session reach your AI at your next prompt. A Team or Enterprise administrator can block mods; Claude web and Desktop chat use the command as text. The mod uses only free company reads. Your own AI plan supplies the brief. It never runs a paid retrieval or approves a change for you. It can offer to share a memory-file note, but notices only the file path; sharing still needs your approval. **Hints while you work.** On your computer, the mod reads your prompts, Claude's last reply and the text Claude writes into files, and compares them with a list of your company's decision titles. When one bears on your work, it adds one short line naming the decision, such as "Annual plan is $30 per seat (accepted 2026-10-21) states $30; the text above states $25." It adds at most one line a turn and three a session, and never the same decision twice in a session. Say “not relevant” and that decision stays quiet for 7 days. Nothing you type, and nothing Claude writes, is sent to OrgSync, stored or logged. Turn it off in `/config` → OrgSync → Hints while you work. In Codex, the same hints come from the plugin's hooks, which Codex asks you to trust once in `/hooks`; they need `node` and send nothing. **Checking work against decisions.** Before it finishes a customer reply, quote, plan or policy, your AI can send OrgSync one or two sentences it writes about the facts it relies on, with no names or pasted text. OrgSync compares them with your company's accepted decisions for free and keeps nothing. ## Home After you sign in, **Home** fits your day on one screen. Under the greeting, one line says what changed since you last looked, as counts: decisions that changed for you, what needs your answer and anything new in Stories. Each count opens what it counts. What you've seen leaves the line; requests stay until you answer them. When nothing is new, the line says so. With the greeting, Olivia, or Oliver if he is your companion, writes a short brief: one or two sentences on what matters for you now, from the records Home already shows you, with each record it draws on linked under it and the time it was written. It is written when you open Home and rewritten only while Home is open and its records have changed, at most once every 45 minutes and five times a day. When it can't be written, Home shows its own summary instead. Below it, one card shows the next thing only you can move, from your company's records; it leaves when you act on it. The card says what kind it is, then its title, one line of what it's about in its writer's own words, who made it and when, and, in bold, what it needs from you. The line can be your own draft's summary, which only you see, a request's purpose, what an AI employee wrote or a problem report. Everyone waiting on you shares one card, which says who is waiting and what it needs from you: people's requests, your AI employees' requests and, for an Owner or Decision approver, suggested corrections and problem reports. Otherwise the card is your own next step, labelled **Your next move**. Its button says where you answer: **Continue in your AI** copies a prompt and names the AI app to paste it into, and copying keeps the card; an AI employee's request opens on this website. **Later** moves the card aside until 9:00 the next morning in your local time, and it stays as the first line under the card until then; choose that line to bring it back. Under **More**, **Not relevant to my work** removes an insight from your Home. When one of your AI employees is waiting for your yes, the card's button reads **Answer now** and the card becomes a review, one request at a time, oldest first. Each shows what the AI employee wrote, marked as its own words, the exact revision your answer applies to and any warning. Choose **Approve**, or **Decline** with a reason the AI employee sees; **Open full** shows the whole request and **Skip** moves on. On a computer, press A to approve, D to decline and → to skip. People's requests are in the same queue with **Continue in your AI**, because you answer them in your AI app. **Back to Home** or Esc ends the review. Under the card is a view by role, with tabs when you have more than one. **Company**, for an Owner or Decision approver, shows goals off track or without a recent check-in, blocked requests, the request that has waited longest and what moved this week. **Team**, for anyone people or AI employees report to, shows each of them with one word from their records, such as **Blocked**, **Waiting on you** or **No check-in**. **Mine**, for everyone, shows what's due today, decisions that changed, who you're waiting on and your key result. Each line opens its page. Billing, data and invitation notices share one line under the collaboration overview. Above the greeting are the story rings, described in [Stories](https://orgsync.co/help#stories), and on the row below them your collaboration overview, described in [Weekly contributions with your team](https://orgsync.co/help#collaboration). On a computer, the menu at the top has **Home**, **Goals**, **Work** and **Files**, and **Settings** is the gear just left of the credit ring. On a phone, all five are in the menu at the bottom. **Goals** groups goals into Company, Your team, Yours and Everyone else, each with its key results and latest check-in, and can show only those due this week, this month or this quarter. Your team is the people who report to you, or, if no one does, your manager and the people under them, as set in **Settings › Team**. **Work** shows requests on a board (To do, Doing, Waiting for review and Done). On a computer the four sit side by side: one is open with whole cards, and the others show each card's title, person and due date. Rest the pointer on one, click it or tab to it to open it. Under **Due**, choose **This week** or **This month** to see work due by then, overdue work included, or **Custom** to pick exact dates; finished work counts by the day it was done. You can also narrow the board by person, goal or tag. Both cost no credit to read and are read-only: each goal and card has **Continue in your AI**, which copies a prompt for your AI app, where changes happen. **Team** and **AI apps** are in **Settings**. **Files** keeps your company's documents in folders. Choose **Upload files** or drop files on the page; they go into the folder you have open. Your AI app can read a file's text and turn it into a draft; images, spreadsheets and presentations are kept as they are. Once that draft is shared, its source names the file it came from and the file shows **Shared as** with its sources, to anyone who may open both. Each folder says who can open it: Owners and anyone who manages a folder choose **Share…** to give everyone, a permission, a team, a person or an AI employee no access, view, edit or manage, and can keep AI apps out of it. A new folder uses the access of the folder it's in until someone sets its own. Each plan includes storage: 5 GB on Standard, 25 GB on Pro and 50 GB on Max. On a paid plan, storage above that is charged from company credit at the rate shown in **Billing**. On Home, **To-do** and **Calendar** sit at the left edge of the screen (on a phone, in the header): point to them or tab to them to see their names, and choose one to open it in a window over the page. **Calendar** shows your deadlines, work in progress, reviews and decisions by month or week: work in progress is a thin line across the days a goal's period runs or a request is under way, and an accepted decision shows on the day it takes effect. Under **Mine** it also shows your dated to-dos. Choose **Mine** or **Company** and **Month** or **Week**, turn kinds of item on or off, and choose a day to list everything on it; until you do, the list beside the calendar shows what is overdue, today and the next 7 days. Each item opens its page or has **Continue in your AI**, and **Download calendar** saves the dates, without to-dos, as a calendar file. Click outside the window or press Escape to close it. To-do and Calendar cost no credit and change nothing. The ring at the top right shows how much of this period's credit is used. Point at it or choose it to see when the credit resets or your free trial ends, open **Billing and usage**, or sign out. Until your AI app is connected, Home shows a **Connect your AI app** card that opens **Settings › AI apps**. Home uses free company reads, though like other free reads it needs a plan or an active free trial. Only its brief uses company credit, at the AI provider's actual cost plus 30% and at most 5% of the period's credit each month. Apart from answering your AI employees' requests, it never shares, approves or changes anything: that work happens in your AI app. ## Stories The story rings at the top of Home are a short catch-up on significant changes at your company. The first ring, Olivia's, is **For you**: updates chosen for your role, responsibilities and active goals, each with a short reason it matters to you. Once your company has team tags, the next rings are one for each team you're on and up to three projects or topics you prioritize, then **Company** for everything else, six rings at most. Until then, the Goals, Decisions, Company, People and Reviews rings appear when your company has updates there. Open a ring to browse more. Each story says where it came from: **Summarized by Olivia** for an AI-written summary, or **From company records**. It names the original authors and links to the company record, which decides the facts and their status. A summary can be wrong, so check the record before relying on it. Stories use only records everyone in your company can read, never private drafts. You can like a story or comment on it. Comments are plain text of up to 1,000 characters, and everyone in your company can read them. You can edit or remove your own; Owners and Decision approvers can hide a comment with a reason. Likes and comments never change a company record. To fix one, choose **Suggest a correction** in a comment's menu, pick the company record and review the prompt in your AI app. The change still needs the usual confirmation and review, as in [Questions and citations](https://orgsync.co/help#asking). In **Settings › Stories**, choose what **Show me** includes. **Only what affects me** shows stories about your goals, requests you sent or received and your work profile, plus what you prioritize. **Important company news**, the default, adds what you follow and news that matters across the company. **Everything** shows every story you haven't muted. A quiet day has fewer stories. Your day resets in your own time zone. You can prioritize, follow or mute each team, project, topic and person, and each category such as Decisions or Goals. Change them in **Settings › Stories** or from a story's **More options** menu. With no setup, you follow the teams you're on, the tags on your goals and on requests you sent or received, your manager, your direct reports and people who mention you in a request. Muting changes only what you see, never who can read a record. Changes that affect you still show. Muted stories leave the rings and For you, and stay in the story list. When you share a decision from your AI app, it can add up to five of your company's tags, such as a team or project, and shows them before you agree. Tags help the right people notice the decision; they grant no access. A correction keeps the decision's tags unless you change them. Writing and choosing stories uses AI, paid from company credit at the AI provider's actual cost plus 30%. That includes attempts the provider bills even when they fail, are rejected or find no stories. There's no charge per story, and reading, liking and commenting cost nothing. Owners can turn automatic stories off or change the monthly stories budget in Settings › Stories. It starts at 30% of the period's included or trial credit, and extra credit doesn't raise it. When the budget or credit runs out, new stories pause and saved ones stay readable. **Settings › Billing** shows the period's Stories AI usage. ## AI employees **Not available to companies yet.** AI employees are being built and tried at OrgSync first. This topic describes how they will work, and details may change before they open to companies. An AI employee is your company's own AI worker in OrgSync. It has its own identity, a role, goals with key results, a monthly credit cap and one responsible human: the person who answers for it, approves what it can't do alone, judges its weekly check-ins and can pause it. It runs Hermes Agent on a computer your company provides and manages, checks for work on a schedule, and submits its results with evidence for a person to review. Everything it creates in OrgSync is labelled as an AI employee's work, with its responsible human. It never uses a person's sign-in or AI app connection. Only Owners add AI employees, up to three per company. Any active member can be a responsible human once they accept. ### The machine - A dedicated computer or virtual machine, used only by this AI employee: no person's accounts, browser profiles, password managers or files. - Linux first: a current Ubuntu LTS virtual machine or server, with about 2 virtual CPUs and 4 GB of memory, more with a desktop for computer use. A dedicated Apple Silicon Mac comes second. Windows isn't supported yet. - Hermes Agent 0.22.5, the version OrgSync tested, with Hermes's own auto-update off. Its gateway runs as a service that restarts after a crash or reboot; on a headless Linux server, turn on linger for its user. - No inbound ports. Hermes's gateway, dashboard and web interface stay on the machine or a private network such as Tailscale. The machine connects out to OrgSync and to its AI providers. - Disk encryption and automatic security updates. On Linux, use your provider's volume encryption so the machine can restart on its own. On macOS, FileVault waits for a person after a restart unless you accept automatic sign-in and its weaker protection. - Your company's own AI accounts, with API keys rather than personal subscriptions, including for Codex CLI or Claude Code CLI if its role needs them. OrgSync never asks for these keys. - Review any skill you add to Hermes from a public hub yourself. Hermes's scan isn't a review. ### Add one and pair its machine 1. **Describe it.** An Owner chooses **Add AI employee** in **Settings › Team**: its name, role, responsible human, who it reports to, how often it checks for work and its weekly check-in day. Nothing is created yet. 2. **Prepare its computer.** The pop-up's last step shows one command. Run it on the machine as the AI employee's own user, in its Terminal or over SSH for a cloud server. The script uses no sudo itself, and the machine never signs in to OrgSync. It installs Hermes Agent 0.22.5 at the exact version OrgSync tested, running Hermes's installer only if its SHA-256 matches, lets Hermes's own setup ask for the model account (the key goes to Hermes, never to OrgSync), starts Hermes's gateway as a service, installs the employee kit after checking its SHA-256 (again on every run, so a changed kit is put back) and adds the `orgsync-employee` command. The kit adds OrgSync to Hermes, turns telemetry off and schedules the work check, the weekly check-in, daily machine care and a heartbeat. At the end it asks for the 8-digit pairing code. You can read the script first, at [orgsync.co/ai-employee/install.sh](https://orgsync.co/ai-employee/install.sh), or run each step yourself under **Do it by hand**. Running it again is safe. 3. **Create it and get the pairing code.** Choose **Create … and get the pairing code** in the pop-up. The AI employee starts with a $10 monthly credit cap, which an Owner can change on its page. The pop-up shows an 8-digit code, such as `4829 1037`, with a Copy button. It works once, for 15 minutes. Type it only on the AI employee's own machine, when the command asks, and don't share it. The machine prints its key fingerprint and names its last 4 characters. 4. **Type the fingerprint's end and pair.** The computer appears in OrgSync by itself, with the setup it declares. The person who answers for the AI employee types the last 4 characters of the fingerprint the machine printed and chooses **Pair and accept responsibility**: in the pop-up when they added it themselves, otherwise on its page or at [orgsync.co/pair](https://orgsync.co/pair). OrgSync never shows the fingerprint, so those characters can only come from the machine. After 3 wrong tries the pairing closes and nothing is paired. If the machine said the code didn't work, choose **Not my computer**: someone else's computer may have used it, so that pairing closes for good. A code that expired or was closed is replaced with **Get a new code** on its page, by the person who answers for it or an Owner; after an Owner's **Replace credential**, only an Owner. Nothing pairs without that click in OrgSync, and a guessed code can't pair a computer, because its fingerprint won't match. 5. **Check it.** Run `orgsync-employee check` on the machine. The first heartbeat moves the AI employee from Setting up to Running. **Finish it later.** If you closed the pop-up or the code expired, open the AI employee from **Not set up yet** under Settings → Team and choose **Get a new code**. On its computer, run `orgsync-employee pair` and type the new code when it asks; the rest of the setup is already done. Running the setup command again also works: it skips what's finished. ### What it can do - **On its own:** read what any member can read, keep private drafts, work on requests it agreed to or was assigned and on its own key results, report blockers, and save its weekly check-in for its responsible human to judge. - **With your yes while Supervised,** the level it starts at: agree to or decline a colleague's request, propose another date, or send a request to a person. At Trusted, which only its responsible human or an Owner can set, it does these alone. - **Always with your yes:** share a draft with the company, suggest a correction, or change its own goals, key results, targets or dates. - **Never:** accept or dispute company decisions, delete sources, assign or review anyone's work, or change billing, members, connections, ways of working or other AI employees. Every piece of work it takes names a request, an assignment or one of its key results. When a task conflicts with an accepted decision or a company policy, or it isn't sure something is lawful, it stops and asks its responsible human. OrgSync's guidance isn't legal advice. ### Approving its requests When it needs your yes, OrgSync records a request and changes nothing yet. You see it in Home's **Only you can** card, and in your AI app's `/orgsync:today`. Open it on the website, signed in, to read exactly what it wrote, marked as written by an AI employee, then choose **Approve**, or **Decline** with an optional reason. You can also answer with OrgSync's prompt in an AI app that shows it. A yes typed in a chat never counts for an AI employee's request, because text it wrote could steer the AI that reads it to you. A request expires after seven days. If the record changes before you answer, the request closes and the AI employee asks again. ### Running, Late and Stopped Its page shows Running while it checks in, Late after two missed work checks and Stopped after six: after 30 and 90 minutes with the usual 15-minute schedule. Home tells its responsible human when it's late, stopped, paused or at its credit cap, or when its work was interrupted. OrgSync sends no email about AI employees. The machine's health figures on its page come from the AI employee itself. A compromised machine could report false ones, so they inform people and grant nothing. ### What OrgSync can and can't control OrgSync controls the AI employee's credential, its records and its work in OrgSync. Its permissions, credit cap, autonomy, rate limits and pause are enforced on OrgSync's servers, and nothing on the machine can change them. OrgSync can't see, switch off, wipe or update the machine, uninstall software or delete files on it. The AI employee looks after its own machine within the rules its kit gives it. Your company is responsible for the machine and the AI services it uses. ### Pause, remove or replace its credential - **Pause** on its page, or ask your AI app. OrgSync refuses its calls except its check-in, which tells it that it's paused, and its held work is released and flagged for you. Resume on the website or with OrgSync's prompt, never with a yes in a chat. - **Replace credential** when you think its machine was compromised. An Owner chooses it on its page: the old credential stops at once, requests it made are closed, and the machine pairs again with a new pairing code and a new key. - **Remove** ends its access for good. Its records stay, labelled as its work, and its open work is flagged for someone else. Then wipe the machine and revoke its AI keys yourself, because OrgSync can't. - If its responsible human leaves, it pauses until an Owner names someone new and they accept. ### Cost There's no seat fee. Its paid calls in OrgSync use company credit at the usual rates: 1¢ per cited answer, and 10¢ per update you approve it to share, once ready for AI. They stop at its monthly credit cap until the next month, in UTC, or until an Owner raises the cap; free reads and its own work continue. **Settings › Billing** shows what each AI employee used this month. Your company pays for the machine, Hermes's AI provider and the coding tools directly. ## Sharing and approving To share an update, ask your AI, for example “Share with *your company*: ” and what changed. Your AI prepares a private draft: a summary and the excerpts that support it. Only you can see your drafts, and nothing is shared until you agree to that exact draft. ### OrgSync's prompt By default OrgSync asks you in its own prompt inside your AI app, which your AI can't answer for you. It starts “OrgSync: share” with the draft's title and your company's name, shows the summary and short previews of the excerpts, and says that sharing publishes that revision to everyone in your company under your name and that your company is charged for one publication once processing finishes. You can ask your AI to show the full excerpts before you answer. - Accept, with its one box ticked, to share that exact revision. If your AI changes the draft, OrgSync asks you again. - Decline, untick the box, or dismiss the prompt to keep the draft private. A declined or dismissed draft can't be shared later; to share it after all, ask your AI to submit it again, and OrgSync asks you again. If your AI app can't show OrgSync's prompt, use [chat confirmation](https://orgsync.co/help#chat-confirmation). Your AI can also list your private drafts and discard one you don't want. A discarded draft is erased and can never be shared. ## Confirm changes in the chat Some AI apps, such as Claude web, can't show OrgSync's prompt. Open your connection in **Settings › AI apps** and, under **How I confirm changes**, choose **In the chat**, then **Save**. Chat confirmation is for when you're in the conversation yourself; for AI that runs on a schedule or without you, keep OrgSync's prompt, where no answer means nothing changes. Only you can change this setting; your AI can't. There's no slash command to change it. Your AI shows the exact draft or change and asks for an explicit yes in the conversation. This covers sharing, changing a knowledge status, accepting or rejecting a suggested correction, setting or changing a goal, and confirming a weekly review. Only an Owner or Decision approver can change knowledge status or decide corrections. Changing the setting costs nothing. Sharing or accepting a correction uses 10¢ once the publication is ready; a status change or rejecting a correction costs no company credit. It never gives your AI extra authority. Deleting a source still requires OrgSync's prompt in an app that can show it. OrgSync checks that the exact change still matches, but can't prove you saw it or said yes. An AI misled by instructions hidden in a document could confirm without asking you. Use OrgSync's prompt when your AI app can show it. A no or no answer leaves the draft private or the change unmade. ## Questions and citations Ask your AI about your company in any chat. It reads the company knowledge you can see and answers with citations. Each citation names the source, its status, when it was shared, its effective date if one was recorded, and its page on OrgSync, which signed-in members can open. That page is read-only. If OrgSync has no source for your question, your AI says so instead of guessing, and offers to ask the colleague who would know through a [request](https://orgsync.co/help#request). Ask what the company decided, and your AI can search accepted decisions only. It can open a cited source in full, or check whether a citation is still current. Answers can also use your colleagues' current roles, for questions about who is responsible for what. ### Statuses - **Reported**: Shared by a member. Every update starts here; it isn't an endorsement. - **Accepted decision**: An owner or decision approver accepted it as a company decision. - **Disputed**: An owner or decision approver marked it disputed. - **Superseded**: Replaced by a newer revision. It stays in the source's history and leaves current answers. - **Withdrawn**: Withdrawn by an owner or decision approver; it's no longer current guidance. “Ready for AI” is separate: it means processing finished, not that the company accepted it. ### When something is wrong - Ask your AI to report a problem with a source: incorrect, out of date, missing context, or something else, with your note. A report changes nothing in the source; owners and decision approvers can list reports from their AI app. - Or ask it to suggest a correction: replacement text with your reason. Once you agree, it's shared with your company's owners and decision approvers under your name, and it changes nothing until one of them accepts it. You can withdraw it until then. - Owners and decision approvers can, from their AI app, accept a source as a company decision, mark it disputed or withdraw it, change an accepted decision's effective date, accept or reject a suggested correction, and delete a source. Status changes and correction decisions need confirmation through OrgSync's prompt or chat confirmation; deletion always needs the prompt. An accepted correction becomes the source's new current revision, and the company is charged for one publication. - The member who shared a source can also delete it while it's still reported information. Deleting is permanent: the source leaves answers right away and its stored content is erased when cleanup finishes. AI chats that already received it keep their copy. ## Plans, free trial and usage One plan covers everyone in your company, with no seat limit. Each plan includes its price as credit every month, shared by everyone; unused monthly credit expires when the month renews. - **Standard**: $25 a month, with $25 of credit each month and 5 GB of company file storage. 7-day free trial with $5 of credit, card required. - **Pro**: $100 a month, with $100 of credit each month and 25 GB of company file storage. 7-day free trial with $15 of credit, card required. - **Max**: $200 a month, with $200 of credit each month and 50 GB of company file storage. No free trial. ### Free trial An owner can start a 7-day free trial of Standard or Pro in **Settings › Billing**. The payment page asks for a card, and nothing is charged when the trial starts. The trial includes its own credit instead of the monthly credit; credit left when it ends expires and has no cash value. When the trial ends, the plan's monthly price, $25 for Standard or $100 for Pro, is charged to that card, and the plan and its monthly credit begin. To stop that, cancel in **Settings › Billing** any time before the trial ends: nothing is charged, and the trial keeps running to its end. OrgSync emails the owner who started the trial 2 days before it ends. There's one free trial per company and per owner. ### Usage Credit is used at 10¢ per shared update once it's ready for AI, and 1¢ per answer that cites company knowledge. Answers without citations are free, and so is opening a cited source. An accepted correction counts as one shared update. Stories use credit at the AI provider's actual cost plus 30%, within a monthly stories budget; see [Stories](https://orgsync.co/help#stories). Home's brief uses credit at the AI provider's actual cost plus 30%, within a monthly brief budget of 5% of the period's credit. Listing decisions, reading your picture and opening a source cost nothing. These free reads still need a plan or an active free trial, but don't need credit left. They don't run a paid cited retrieval. Your AI app's own plan pays for its work on imports, interviews and briefs. Without a plan or an active free trial, your AI can't read company knowledge. Shared updates wait until an owner starts one; if the company runs out of credit during it, they wait until there's credit again. ### Extra credit, cancelling and refunds - Owners can add credit in **Settings › Billing**, from $10. It's used after monthly credit and carries across renewals and into a later subscription of your company. Each purchase expires 6 calendar months after you buy it, and using it needs a plan or free trial. - Cancel renewal at any time in **Settings › Billing**. It takes effect at the end of the paid month, and access continues until then. - Neither the monthly price nor extra credit is refunded in cash. - After paid access ends, you have 30 days to export company content or restart before it's deleted. Prices are in US dollars; the payment page shows the final amount, including any tax, before you pay. **Settings › Billing** shows the plans, rates and trial credit that apply to your company before you pay. If anything here differs, Billing is right. ## Your data - **Export:** anyone in the company can download an export in **Settings › Data**: everything they can see, with its history. The ZIP has one file for each source, goal and review, the company files they can open, and the same records as JSON; a single JSON or Markdown file suits a small company. Someone else's private drafts are never included. - **Delete a source:** from your AI app, as in [Questions and citations](https://orgsync.co/help#asking). - **Remove someone:** owners do this in **Settings › Team**. Their future requests are refused and their AI apps' connections stop working. What they already shared stays. - **Disconnect an AI app:** in **Settings › AI apps**, open the connection and choose **Disconnect**. It stops working with your company from its next request. - **Delete the company's content:** owners do this in **Settings › Data**, after cancelling the plan's renewal. It removes shared knowledge, drafts, processed copies and roles, and blocks access right away. Backups that still contain it expire later; Settings › Data shows when. To process shared updates, answer questions and draft roles in the onboarding chat, OrgSync sends that text to the AI model providers it uses. It doesn't send private drafts until they're shared. OrgSync can't erase or recall copies that someone already exported or that an AI app already received. Where OrgSync keeps your data, who processes it and how it's protected are on [Security and data handling](https://orgsync.co/security). ## Troubleshooting ### My connection doesn't show in AI apps After you sign in, the connection shows **Finish connecting** until your AI checks it: ask your AI “Check my OrgSync connection.” AI apps keeps looking for up to 10 minutes while it's open; then choose **Check again**. If nothing appears, make sure you chose **Allow** on the OrgSync page that opened in your browser, and that it showed the right company. In Codex, use the sign-in command from step 2 and cancel any sign-in Codex started by itself. On a Claude Team or Enterprise plan, a workspace owner adds the connector first. An expired connection needs connecting again. ### My AI says it needs a plan or credit If it says a plan is required, ask an owner to start the free trial or subscribe in **Settings › Billing**. Adding credit alone doesn't start paid access. If the plan or free trial is active but credit has run out, an owner can add credit there, from $10. Shared updates wait meanwhile and continue once there's credit. Free reads can still run with no credit left while access is active. If credit couldn't be checked, try again later. ### OrgSync's prompt didn't appear If your AI app can't show OrgSync's prompt, the draft stays private. Choose **In the chat** for that connection, as in [Confirm changes in the chat](https://orgsync.co/help#chat-confirmation), then ask your AI to share the draft again. If you dismissed or declined the prompt, ask your AI to submit the draft again, and OrgSync asks you again. If you don't answer, nothing is shared. Status changes and correction decisions also support chat confirmation. Deleting a source needs an app that can show OrgSync's prompt and a connection set to use it. ### I can't delete a source from ChatGPT or Claude web Deleting a source needs OrgSync's prompt, which ChatGPT and Claude web can't show, and chat confirmation doesn't cover it. ChatGPT may also stop the request before it reaches OrgSync. Delete it from an app that shows OrgSync's prompt, such as the Claude Code CLI, on a connection set to use the prompt, or ask an owner. In the chat, an owner or decision approver can withdraw the source or mark it disputed instead, which takes it out of current guidance without deleting it, and anyone can report a problem with it. ### Claude Code uses a different OrgSync connection If you added OrgSync to Claude Code yourself before installing the plugin, `/mcp` lists that server instead of **plugin:orgsync:orgsync**: when both point to the same address, Claude Code uses the one you added. Its tools have the same names whichever company it signed in to. To use the plugin's connection, run `claude mcp remove orgsync` in Terminal, with the name `/mcp` shows, then restart Claude Code and sign in to **plugin:orgsync:orgsync** from `/mcp`. To check which person and company a connection uses, ask your AI “Check my OrgSync connection.” or run `/orgsync:today`. ### My connection expired or was disconnected Open **Settings › AI apps** in OrgSync and connect your AI again. A disconnected connection can't be restored; connecting again creates a new one. If your company access ended, ask an owner for help. ### I need to sign in again Sign in again on OrgSync to continue. If you're verifying an email address, open the link in your email, then sign in. For an expired or invalid link, sign in with the same email and password to get a new one when email verification is available, or continue with Google using that address. ### My company's content is being deleted Deletion blocks reading, sharing and exporting company content. An owner can check progress in **Settings › Data**. Adding credit or reconnecting your AI doesn't undo deletion; ask the owner if you weren't expecting it. ### My invitation link doesn't work It works only for the email it was sent to, so sign in with that one. If it expired, was cancelled or was already used, ask the owner for a new link. ### OrgSync's new commands or tools don't show up After OrgSync adds or changes a command or tool, your AI app can keep its old list. Use the step for your app: - **Claude Code:** in Terminal, run `claude plugin marketplace update orgsync`, then `claude plugin update orgsync@orgsync`, then restart Claude Code. - **Claude Desktop:** open **Settings › Plugins › OrgSync**. It shows your installed version and an **Update** button when a newer version exists. - **claude.ai:** the OrgSync marketplace can stay on an older version until you choose **Check for updates** for it. - **ChatGPT:** open the OrgSync connection in ChatGPT's settings and choose **Refresh tools**. ### Claude Desktop says “Couldn't check for updates” The Code tab in Claude Desktop can show that message. It comes from Claude Desktop itself, not from OrgSync. Open **Settings › Plugins › OrgSync**: if it shows the latest version, nothing is wrong. The latest version is listed in the plugin repository, `orgsync-ai/plugins`. ### ChatGPT warns about “suspicious content” When you say yes to a change in the chat, ChatGPT's own permission prompt may call the confirmation code that your AI then sends “suspicious content”. That code is the one-time code OrgSync issued for the exact change you just approved. Allow it only if you just said yes to that change. Otherwise decline it; nothing changes. ## What's new New and changed features, newest first. Your AI app reads the same list in the text copy of this help, so you can ask it what's new in OrgSync. - **October 30, 2026**: Home's brief is now written by Olivia or Oliver: one or two sentences on what matters for you now, from the records Home already shows you, with each record it names linked under it. It is rewritten only while Home is open and something changed, and uses company credit at the AI provider's cost plus 30%. - **October 29, 2026**: Company files: Files in the menu keeps your company's documents in folders. Upload PDFs, Word, Excel, PowerPoint, text, Markdown, CSV and images. Each plan includes storage: 5 GB on Standard, 25 GB on Pro and 50 GB on Max; Owners see how much is used in Billing. Owners and folder managers choose Share… to set who can view, edit or manage each folder, and whether AI apps can read it. Your AI app can read a file's text and turn it into a draft. - **October 25, 2026**: Inviting someone now also asks what the role is for and what they decide alone or ask about. A new member sees Your first weeks on Home for their first 45 days: their role, who they report to, their AI app, a first goal and a first weekly review, each ticked once it is done. Their manager sees how far they have got on the team strip and the Team page. - **October 24, 2026**: Work's four columns now sit side by side on a computer. One is open with whole cards; the others show each card's title, person and due date, and open when you rest the pointer on them, click them or tab to them. - **October 24, 2026**: To-do and Calendar now sit at the left edge of Home (in the header on a phone): choose one to open it. To-do lists your own to-dos for this week, this month or all of them, overdue first, with each one's note, repeat, reminders and links; only you can see it, and Continue in your AI changes one. Calendar adds a Week view, thin lines for work in progress, decisions on the day they take effect and, under Mine, your dated to-dos. A downloaded calendar and your AI app's agenda no longer show an agreed date once a request's terms have changed. - **October 23, 2026**: Home has a Widgets tab at the left edge of the screen (a Widgets button on a phone). It opens Calendar over the page: your deadlines, reviews and decisions by month, Mine or Company, with what is overdue, today and the next 7 days beside it. Each item opens its page or copies a prompt for your AI app, and Download calendar saves the dates. - **October 23, 2026**: Work is now one board with a Due filter: All, This week, This month (overdue work included) or Custom dates, and the address keeps your choice. On Goals, Any time is now All, and a link to a goal opens Goals at that goal. - **October 23, 2026**: Home, Goals, Work and Settings now share one page width. Work's board shows its four columns two by two, and Home's story rings sit centred above wider cards. - **October 23, 2026**: Ask your AI app to remember things for your work: run `/orgsync:todo` for a private to-do with a date, time or repeat. Due to-dos show in `/orgsync:today`, and a follow-up can close itself when a teammate submits. OrgSync never emails to-dos. - **October 23, 2026**: On a computer the menu now sits at the top beside the logo, and pages slide as you move between them. Goals groups goals into Company, Your team, Yours and Everyone else and can show only those due this week, month or quarter. Work adds a month calendar beside the board. Card buttons now read Continue in your AI, and the credit ring opens a menu with Billing and Sign out. - **October 23, 2026**: `/orgsync:today` now opens with one ranked list of what to do: what someone is waiting on first, then what is due, work for a company goal and the priorities from your weekly review. - **October 23, 2026**: Stories now asks what to show you instead of how many: only what affects you, important company news or everything. Follow, prioritize or mute teams, projects, topics and people in Settings › Stories or from a story's menu; changes that affect you still show. Once your company has team tags, the rings follow your teams. Your AI app can add tags when you share a decision. - **October 23, 2026**: Ask your AI app in your own words and it starts the right OrgSync step. New: `/orgsync:request` to ask or assign a teammate, `/orgsync:meeting` to turn meeting notes into decisions and follow-ups, and `/orgsync:team` to see how the team is doing. `/orgsync:get-started` now fits your role, and bringing your company in can look through your past chats for decisions, with your yes. Update the OrgSync plugin to 0.13.0 to get them. - **October 23, 2026**: The bottom menu is now Home, Goals, Work and Settings. Goals shows each goal's key results and latest check-in; Work shows requests on a board from To do to Done. Copy a prompt from any card to make a change in your AI app. Team and AI apps moved into Settings. - **October 23, 2026**: After you connect your AI app, run `/orgsync:get-started`: a short welcome, three quick choices, then help bringing your company in and setting your goals. - **October 22, 2026**: Bring your company in now shows how much your AI can draw on across eight areas, and what's still missing. - **October 22, 2026**: Team shows who reports to whom. Owners can choose Arrange, drag people under their new manager and save the whole change at once. - **October 22, 2026**: Prompt buttons now copy with one click and tell you which AI app to paste into. - **October 22, 2026**: The setup guide now opens each time you come back until setup is done, or until you choose Later on its Home card. - **October 21, 2026**: Help is organized by topic, with search, and now covers Home, Stories and key results. Your AI app can read it as text at orgsync.co/help.md. - **October 20, 2026**: Creating a company takes three short steps. A setup guide then helps you choose a plan, describe your role and connect your AI app. - **October 19, 2026**: Goals can have up to three key results, and your weekly review records their values. Owners turn OKR and Kanban guidance on or off in Settings › Company, set a work-in-progress limit and keep a list of tags. - **October 19, 2026**: Weekly contributions with your team: from your AI app, ask a teammate for information or work toward a goal, agree its terms, report blockers and review the result. Home shows an overview, and you can download the agenda for your calendar. - **October 18, 2026**: Stories on Home: a short daily catch-up on significant changes at your company, chosen for your role and goals, with likes and comments. Choose how many you get in Settings › Stories. - **October 18, 2026**: Home cards are fitted to your role, with prompts you can edit and copy into your AI app. - **October 18, 2026**: The ring in the header shows how much of this period's credit is used. Choose it to open Billing. - **October 17, 2026**: `/orgsync:guide` helps you choose one useful next goal, starting from what your company has agreed. - **October 15, 2026**: `/orgsync:review` runs your weekly review, with an optional reminder email. ## Contact Questions? Email [support@orgsync.co](mailto:support@orgsync.co) (Mon–Fri, 9:00–17:00 Toronto time). Say what you were doing and any message you saw. Never send a password or a connection key. How OrgSync handles personal information is in the [Privacy notice](https://orgsync.co/privacy); the rules for using it are in the [Terms of service](https://orgsync.co/terms). Where your data is kept, who processes it and how to report a vulnerability are on [Security and data handling](https://orgsync.co/security).