--- title: "Use Notion Docs With AI Agents (and Confluence)" url: https://www.insulin.dev/blog/notion-and-confluence-docs-in-insulin/ canonical: https://www.insulin.dev/blog/notion-and-confluence-docs-in-insulin/ type: Blog description: "How to use Notion docs with AI agents, and Confluence: what an Insulin knowledge base can index today, what agents reach live, and which agents can." --- # Use Notion Docs With AI Agents (and Confluence) > Canonical HTML version: https://www.insulin.dev/blog/notion-and-confluence-docs-in-insulin/ 1. [Home](/) 2. / 3. [Blog](/blog/) 4. / 5. Use Notion Docs With AI Agents (and Confluence) # Use Notion Docs With AI Agents (and Confluence) Notion isn't a knowledge-base connector yet and Confluence has none, but agents can reach both live. Which agents can, what they can change, and when to index instead. ![Shirley Guo](/authors/shirley-guo.jpg) Shirley Guo Oct 7, 2026 ![Use Notion Docs With AI Agents (and Confluence)](/images/blog/notion-and-confluence-docs-in-insulin/hero.png) Explore AI Summary [![](/logos/company/openai.svg)](https://chat.openai.com/?q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fnotion-and-confluence-docs-in-insulin%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Integrations%2C%20Knowledge%20Bases. "Summarize with ChatGPT")[![](/logos/company/anthropic.svg) ](https://claude.ai/new?q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fnotion-and-confluence-docs-in-insulin%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Integrations%2C%20Knowledge%20Bases. "Summarize with Claude")[![](/logos/company/gemini.svg)](https://www.google.com/search?udm=50&aep=11&q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fnotion-and-confluence-docs-in-insulin%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Integrations%2C%20Knowledge%20Bases. "Summarize with Gemini")[](https://www.perplexity.ai/search/new?q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fnotion-and-confluence-docs-in-insulin%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Integrations%2C%20Knowledge%20Bases. "Summarize with Perplexity") Table of Contents - [Can an Insulin knowledge base sync Notion or Confluence?](#can-an-insulin-knowledge-base-sync-notion-or-confluence) - [How do agents reach Notion live?](#how-do-agents-reach-notion-live) - [How do agents reach Confluence live?](#how-do-agents-reach-confluence-live) - [Index it or fetch it live?](#index-it-or-fetch-it-live) - [What changes between an indexed answer and a live one?](#what-changes-between-an-indexed-answer-and-a-live-one) - [Which control matters on each route?](#which-control-matters-on-each-route) - [Frequently asked questions](#frequently-asked-questions) - [Takeaways](#takeaways) _An Insulin knowledge base can’t sync Notion or Confluence today: the documentation says the Notion connector isn’t available yet and commits no date, and there is no Confluence connector. Agents reach both live instead, through the Notion and Atlassian integrations, and who owns the connection decides which agents can._ * * * “Our docs are in Notion. Can Insulin answer from them?” It is one of the first questions an evaluator asks, and for a Confluence team it is the same question with a different logo. The honest answer starts with a distinction, because documentation reaches an agent in two ways, and Notion and Confluence each support only one of them today. **Indexing is copying documents into a knowledge base, where an agent searches them and cites the source document for each result.** **Fetching live is an agent calling an integration’s tools against the system itself, as it works** — and for both Notion and Confluence those tools write as well as read. What follows: where each kind of documentation can go, which agents can use each route, and the one control that matters on each. ## Can an Insulin knowledge base sync Notion or Confluence? **No, not today.** A knowledge base holds three source types — uploaded or pasted files, crawled websites and synced connectors — and the connectors are GitHub, ClickUp and Google Drive. The [knowledge base documentation](https://doc.fours.com/insulin/knowledge-base/) says Notion isn’t available yet and commits no target date. Confluence isn’t among the connectors at all. **A connector is a synced source that keeps a knowledge base in step with one of your connected integrations.** With **Auto-sync** on, Insulin sweeps it about once an hour. Which connectors a knowledge base takes depends on who owns it: - **Organization knowledge bases** support only the Google Drive connector. - **GitHub and ClickUp** connectors are available on user knowledge bases. Setting those connectors up, and reading each source’s sync status, is covered in [how to connect Google Drive to an AI knowledge base](/blog/connect-external-sources-to-a-knowledge-base/). Notion and Confluence pages can still reach a knowledge base, but only as copies: - **Upload an export.** A knowledge base reads PDF, DOCX and text formats including Markdown and HTML, up to 25 MB per file. Uploads don’t auto-sync, so the copy stays as it was on the day you exported it. A same-name upload replaces the existing document in place, so a fresh export is a clean swap. - **Crawl what is already public.** The Websites tab takes a public URL and a depth from 1 to 10, and imports up to 100 pages. It re-crawls only when you run **Re-sync**, and internal and private hosts can’t be crawled, so a wiki on an internal address is out of reach. ## How do agents reach Notion live? **Through the Notion integration, which gives an agent tools rather than an index.** The [Notion integration page](https://doc.fours.com/integrations/notion/) counts 45 actions in six groups: pages, databases, database queries, blocks, comments, and search and users. They search the workspace, read a page or export it as Markdown, and run filtered or sorted database queries; they also create, update, duplicate, move and archive pages, update databases and their schema, manage blocks and post comments. Notion connects at two levels, and the level decides which agents can use it: Connection How it is made Which agents can use it **Organization** A token from a Notion internal integration, shown as “Notion (API Key)” and shared across the organization. It reaches only the pages and databases shared with that integration in Notion Organization agents with it on their allowlist, because they can use organization-level integrations only **Personal** Each person connects their own Notion workspace through OAuth Your personal agents, when you select it for them, and the built-in Insulin assistant, which runs with your connected accounts Where both levels are connected, the integration page says the user-level connection takes precedence. An organization agent never makes that choice: it can’t use a personal connection at all, so the organization token is its only way into Notion. ## How do agents reach Confluence live? **Through the Atlassian integration, which is user-level only.** Each person connects their own Atlassian account through OAuth, and the [Atlassian integration page](https://doc.fours.com/integrations/atlassian/) counts 78 actions across Jira and Confluence. On the Confluence side, an agent can create, update and search pages and blog posts, add footer and inline comments and content properties, read spaces and manage attachments, across the Atlassian sites your connected account can access. Two documented statements decide who can use it: the Atlassian integration is configured per user, with no organization-level option, and organization agents can only use organization-level integrations. Put together: - **The built-in Insulin assistant and your personal agents can reach Confluence**, through your own Atlassian connection. - **No organization agent can.** There is no organization-level Atlassian connection for it to use. - **No organization channel can either.** Its only agent members are organization agents; its coordinator, Planner, has no integrations by design; and the Insulin assistant isn’t available there. So when an organization agent has to work from Confluence content, the route is a copy: export the pages and upload them to an organization knowledge base attached to that agent. ## Index it or fetch it live? **Index whatever a knowledge base can sync or hold; fetch live only what it can’t, or when the agent has to act on the current page.** Where the docs are Can a knowledge base index it? Can an agent reach it live? What to do **Google Drive** Yes. It is the only connector organization knowledge bases support Not needed for grounding Connect the folders or files, with Auto-sync on **GitHub issues and pull requests, ClickUp tasks and Docs** Yes, as connectors on user knowledge bases Not needed for grounding Connect them to a user knowledge base, for your personal agents and the built-in assistant **Notion** Not as a connector: not available yet, with no date committed Yes: the organization connection for organization agents, a personal one for your own agents and the built-in assistant Fetch it live from the agents that need it; upload an export where answers must be cited **Confluence** Not as a connector: there isn’t one Yes, but user-level only: the built-in assistant and your personal agents, never an organization agent Fetch it live from your own agents; for an organization agent, upload an export **A public help centre** Yes, as a website: a public URL, depth 1–10, up to 100 pages Not needed for grounding Crawl it, and run Re-sync when it changes **An intranet on an internal host** No. Internal and private hosts can’t be crawled Only if the system behind it has an integration Export the pages and upload them **Exports from any of these** Yes, as uploads: PDF, DOCX, Markdown, HTML and other text formats, 25 MB each Not applicable Upload them, and re-upload under the same name when the source changes For Notion and Confluence, the choice turns on one question: does the answer need a citation, or does the agent need the page as it is right now? A policy an agent should quote and point to belongs in a knowledge base, even as an export you refresh by hand. A page an agent should check before acting, or update as part of its job, is a live fetch. ## What changes between an indexed answer and a live one? **An indexed knowledge base returns passages with the source document cited; a live integration acts with whatever its tools allow.** Indexed knowledge base Live integration **What the agent works from** Passages retrieved from indexed documents, from up to three knowledge bases per turn Tool calls against Notion or Confluence itself **How you check an answer** The source document is cited for each result Each tool call expands to show its inputs and outputs **How current it is** As of the last sync: Auto-sync connectors are swept about hourly, while uploads and websites never update on their own Whatever the system holds when the tool runs **What it can change** In Read mode, nothing; in Edit mode, the knowledge base’s files (create, update and deprecate, never delete) Whatever the integration’s actions allow, which for Notion and Confluence includes creating and updating pages ## Which control matters on each route? **On the indexed route it is the access mode; on the live route it is the integration allowlist.** Both are set on the agent, not written into its prompt. **Indexed: attach the knowledge base in Read mode.** **Read** lets an agent retrieve and search for grounded context only. **Edit** adds creating, updating and deprecating files in that knowledge base, never deleting; keep it for an agent whose job is maintaining the base. **Live: grant the integration only where the job needs it.** An agent’s integrations are an allowlist of connected systems: Notion is on it or it isn’t. With no documented per-action read-only switch, an agent you give Notion to read pages also gets the tools to create, update and archive them, and one with Atlassian gets the tools to create and update Confluence pages. [Scoping an agent to its integrations](/blog/scope-an-agent-to-its-integrations/) covers how to draw that boundary. The built-in assistant is the exception: it can’t be edited and runs with your connected accounts, so for it the control is whether you connect Notion or Atlassian personally at all. To check what an agent was granted, ask it to list the actions an integration supports; each is marked read-only or not. **Don’t count on an approval card for the writes.** Some tool calls pause for your approval, but the documentation doesn’t say a Notion or Confluence write is one of them. Where a call does pause, the card is offered only in an interactive solo chat in the web workspace, on the main thread. In a channel, a Slack or Teams DM, or a sub-thread, it isn’t offered, and the action isn’t held back; [where the approval gate is armed](/blog/where-the-approval-gate-is-armed/) goes surface by surface. **For the Notion organization connection, one more lever sits in Notion.** The token reaches only the pages and databases shared with its integration, and Notion’s own [documentation on connection capabilities](https://developers.notion.com/reference/capabilities), which now calls integrations connections, adds two details: access to a page extends to the pages beneath it, and reading, updating and inserting content are separate capabilities that Notion enforces, so an integration given only Read content can read pages but not change them or create new ones. Share the narrowest parent page that covers the job. ## Frequently asked questions ### Can an Insulin knowledge base sync Notion? Not yet. The documentation says the Notion connector isn’t available and commits no date. Today an agent reaches Notion live through the Notion integration, or you upload exported pages to a knowledge base as files, which don’t auto-sync. ### Can an organization agent read Confluence? No. The Atlassian integration is user-level only, and organization agents can only use organization-level integrations. The built-in Insulin assistant and your personal agents can reach Confluence through your own Atlassian connection; for an organization agent, upload exported pages to a knowledge base. ### Which Notion pages can an organization connection reach? Only the pages and databases shared with the Notion internal integration whose token was connected. A personal connection is separate: each person connects their own workspace through OAuth, and where both levels are connected, the user-level connection takes precedence. ### If an agent can read Notion, can it also edit pages? Yes, on the Insulin side. The allowlist works per connected system, with no documented per-action read-only switch, and the Notion actions include creating, updating and archiving pages, so an agent with Notion on its allowlist gets those tools. Grant Notion only to agents whose job needs it. ### Will Insulin ask before an agent writes to Notion or Confluence? Don’t count on it. The documentation doesn’t say these writes pause for approval, and the approval card appears only in an interactive solo web chat on the main thread. In a channel, a Slack or Teams DM, or a sub-thread, the turn proceeds without pausing. ### Should we index our docs or let agents fetch them live? Index what agents must cite: a knowledge base returns passages with the source document cited. Fetch live when an agent needs the current page or has to act on it, and scope those agents tightly, because the Notion and Atlassian tools can write. ## Takeaways - No Insulin knowledge base syncs Notion or Confluence today: the Notion connector isn’t available yet, with no date committed, and there is no Confluence connector. - Notion reaches agents live at two levels: organization agents use the organization connection; your own agents and the built-in assistant use a personal one. - Confluence reaches agents only through the user-level Atlassian integration, so no organization agent or organization channel can use it. - Live tools can write. The allowlist is the control, set per connected system with no documented read-only switch, and the approval card is not a guarantee. - When an answer must be cited, or an organization agent needs Confluence, upload an export to a knowledge base and replace it when the source changes. To see what a knowledge base holds and how agents cite it, start with [AI knowledge bases in Insulin](/knowledge-bases/). ## Sources Primary sources for the platform rules cited above. Last verified October 7, 2026. Cloud providers change fees, eligibility, and program terms without notice — check the source before relying on a figure. - [Fours Insulin docs: Knowledge Base](https://doc.fours.com/insulin/knowledge-base/) — The three source types and the three connectors (GitHub, ClickUp, Google Drive); Notion not yet available, with no target date committed; organization knowledge bases supporting only the Google Drive connector, with GitHub and ClickUp on user knowledge bases; Auto-sync connectors swept about once an hour, while websites and files never sync on their own; uploads in PDF, DOCX and text formats including Markdown and HTML, 25 MB per file, a same-name upload replacing the document in place; website crawls of a public URL at depth 1–10, up to 100 pages, re-crawled only on Re-sync, with internal and private hosts refused; which knowledge bases user-level and org-level agents attach; the built-in Insulin assistant searching your own knowledge bases by default; agents searching up to three knowledge bases per turn and citing the source document for each result - [Fours docs: Notion integration](https://doc.fours.com/integrations/notion/) — Organization-level access through a Notion internal integration token, shown as Notion (API Key) and shared across the organization, reaching only the pages or databases shared with that integration; user-level access per user through OAuth; 45 actions in six capability groups, including creating, updating, duplicating, moving and archiving pages; both levels supported, with user-level taking precedence when both are connected - [Fours docs: Atlassian integration](https://doc.fours.com/integrations/atlassian/) — User-level only, each user authenticating their own Atlassian account through OAuth; 78 actions across Jira and Confluence; Confluence pages and blog posts created, updated and searched, footer and inline comments and content properties added, spaces read and attachments managed; multi-site access resolved through the connected account's accessible resources - [Fours Insulin docs: Agents](https://doc.fours.com/insulin/agents/) — The built-in Insulin agent unable to be edited; org-level agents able to use only org-level integrations; integration access by agent type; the integrations field as an allowlist of connected systems; the Read and Edit knowledge-base access modes, Edit never deleting; an agent listing the actions an integration supports and whether each is read-only; tool calls expanding to show inputs and outputs; some tool calls pausing for approval, with the card offered only in an interactive solo chat in the web workspace on the main thread, and not in a channel, a Slack or Teams DM or a sub-thread - [Fours Insulin docs: Channels](https://doc.fours.com/insulin/channels/) — Organization channels admitting only org-level agents as members; Planner having no integrations, by design; the built-in Insulin assistant not available in organization channels, because it runs with your connected accounts ## Keep reading - [Knowledge BasesConnect Google Drive to an AI Knowledge BaseAug 18, 2026](/blog/connect-external-sources-to-a-knowledge-base/) - [Knowledge BasesWhich Documents Can an AI Knowledge Base Read?Aug 4, 2026](/blog/what-documents-an-ai-knowledge-base-can-read/) - [IntegrationsEWS Retirement and Your AI Email AssistantOct 8, 2026](/blog/ews-retirement-ai-email-assistant/) - [SecurityRead-Only Access for AI Agents: Where It's SetOct 8, 2026](/blog/read-only-access-for-ai-agents/) [Browse every post on the Insulin Blog](/blog/) ### Stay Updated New posts, product updates and marketplace strategy are shared on LinkedIn as they publish. [Follow Fours on LinkedIn](https://www.linkedin.com/company/suger-inc)