Find the relevant pages
Search across accessible spaces, then use page titles, space keys, and slugs to choose what the assistant should read.
MCP FOR AI ASSISTANTS
Bring your AI.
Raccoon Page brings the company context.
Keep working in the assistant you already use. Connect Raccoon Page through Model Context Protocol so it can search the spaces you can access, read the relevant pages, answer with that context, and update the wiki when your client allows the write.
MCP access is included on every Raccoon Page plan. Any AI client or model charges are separate.
Your AI harness
Model, reasoning, current task, approvals
- Code and local files
- Other connected tools
- Client-controlled write approvals
Raccoon Page
Pages, spaces, permissions, organization
- Search and page retrieval
- Page, label, and comment tools
- Existing user access checks
THE PRODUCT BOUNDARY
Your wiki does not need another chat box.
A page-level rewrite button starts in the wiki. Your daily AI assistant already has the conversation, the repository, the current task, and any other tools you connected. MCP lets the wiki contribute its part without trying to replace that working environment.
Raccoon Page remains the durable source of truth. Your harness remains responsible for model choice, reasoning, and its configured approval policy. That means no second model interface to learn and no Raccoon Page model-usage charge layered on top of the assistant you already pay for.
If your team wants a one-click AI writer embedded in every page with no client configuration, Raccoon Page is not that product today. This workflow is for teams that already work with an MCP-compatible assistant and want the wiki available there.
WHAT IT CAN DO
From a question to a reviewed wiki change.
The assistant calls small, explicit tools. It retrieves the material it needs instead of receiving a bulk export of your workspace.
Read durable context
Retrieve page content as Markdown, walk a page tree, and inspect child pages without scraping the browser interface.
Propose and apply changes
Create draft pages after review, update pages with an edit summary, move pages, and keep labels organized.
Work through comments
List, add, reply to, resolve, or remove comments when documentation work needs a visible discussion.
ONE PRACTICAL PROMPT
Ask across the wiki. Keep the write separate.
This is the useful part of cross-space retrieval: the assistant can search several accessible spaces, combine those pages with the task in front of you, and show its work before anything changes.
prompt
Search the Product, Engineering, Security, and Support spaces for information about session expiry. Summarize the current policy, name the pages you used, compare it with this implementation, and propose an update to the authentication policy page.Do not change the wiki until I approve the draft.
WHAT CONTEXT MEANS
Relevant pages, retrieved when needed.
Raccoon Page currently uses PostgreSQL full-text search with typo matching on page titles and labels. The assistant searches, selects results, reads the useful pages, and performs the synthesis in your chosen model. It does not receive a standing copy of every document.
Access follows the connected user
The personal access token maps to a Raccoon Page user. Existing space and write permissions remain in force.
Use one expiring token per client
Keep tokens out of checked-in files, give each connection a recognizable name, and revoke it when the client is no longer used.
Review first, write second
Request a proposed edit before authorizing an update. Tell the assistant to create new pages as drafts until a person has reviewed them.
USE YOUR CURRENT TOOL
One remote connection. Your choice of assistant.
Raccoon Page uses remote Streamable HTTP with bearer-token authentication. Client interfaces change, so the setup guide remains the canonical configuration reference.
CodexUse the shared MCP configuration from the desktop app, CLI, or IDE extension.Claude CodeConnect the remote HTTP endpoint and send the personal access token as a bearer header.CursorAdd the remote server globally or for one project, with the token supplied from an environment variable.VS CodeConfigure a remote HTTP MCP server in mcp.json and keep the token out of the checked-in file.
Questions before you connect.
Does Raccoon Page include its own AI model?
No. Raccoon Page supplies the wiki tools and accessible content. Your chosen assistant and model perform the reasoning, so their subscription, data handling, and approval controls still apply.
Does connecting MCP expose every page?
No. The connection acts as the Raccoon Page user who created the personal access token. Search and read operations are limited to spaces that user can access, and write operations still pass the product's permission checks.
Can an assistant update the wiki?
Yes, when the connected user has permission and the client's approval policy allows the write tool. MCP does not require a human approval step, so a safer client configuration is to request an answer or proposed edit first, inspect it, and approve the specific change separately.
Which Raccoon Page plans include MCP?
MCP access is included on Free, Team, and Business. Your AI provider may charge separately for its subscription or model usage.
Keep the AI. Connect the wiki.
Start with the Free plan, create an expiring personal access token, and make Raccoon Page available where your team already asks questions and gets work done.