Sign inBook demoStart for free

Slite MCP

Connect your AI assistant to Slite using the  Model Context Protocol (MCP)  — an open standard that lets AI tools interact with external services.
One connection gives your AI assistant the ability to search your knowledge base and connected sources, and to read, write, and organize docs — all from inside Claude, Cursor, ChatGPT, or any MCP-compatible tool.


What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude, ChatGPT, Cursor, or Gemini connect to external tools and data sources. Instead of copy-pasting context into a chat window, your AI assistant can directly read from and write to connected services.
Slite has one MCP that covers both search and writing — no need to choose between integrations.


What can I do with the Slite MCP?

The Slite MCP can…
Search across your docs and return answers
Search across all your connected data sources simultaneously
Draft, edit, and organize docs directly in your workspace
Move, archive, verify, and restructure docs
Upload files from your machine straight into a doc
Do all the above via one MCP, one connection


Why use the Slite MCP?

Get instant answers from across all your connected data sources

Instead of searching through multiple platforms yourself (Google Drive, Slack, Linear, etc.), ask your AI assistant and it'll search across all your connected data sources simultaneously to find the answers you need.
Your team spends less time platform-hopping and more time using information.

Turn existing knowledge into reusable content

Pull information from across your workspace to create new onboarding guides, customer announcements, or internal reports.
What used to take hours of research and writing now takes minutes.

Work where you already work

Interact with your Slite workspace directly from your favorite AI assistant — no context switching, no copy-pasting.

Example prompts

These are example prompts you could send to your AI assistant once the Slite MCP is connected:
  • "Summarize our Q1 product updates and draft a customer announcement."
  • "What's our policy on remote work?"
  • "Draft an onboarding guide using our internal discussions and meetings."
  • "Upload this chart and add it to our weekly metrics doc."


Setting it up

The Slite MCP uses OAuth authentication — no API keys or config files needed. Just add the following URL as a custom MCP connector in your AI assistant's settings:
https://api.slite.com/mcp

The Slite MCP works with Claude, ChatGPT, Cursor, Gemini, Codex, and any other tool that supports the Model Context Protocol.
With Claude Desktop
    Go to Settings > Connectors > Add custom connector , then enter the Slite MCP URL above as well as a name. The advanced settings can stay blank.
    Sign in to your Slite account and authorize Claude.
With ChatGPT/Codex
    In the web app, go to Settings > Apps > Advanced Settings , and enable Developer Mode .
    Create a new app with the Slite MCP URL: https://api.slite.com/mcp
    Select the OAuth authentication option and leave the other fields blank.
    Sign in to your Slite account and authorize ChatGPT.
With other MCP clients
Add https://api.slite.com/mcp as a remote MCP server URL and complete the OAuth authentication when prompted.
With clients requiring manual credentials (e.g. Google Gemini Enterprise)
Some clients do not support automatic Dynamic Client Registration (DCR) and require you to manually provide a Client ID, Client Secret, and OAuth endpoints.
    Register a client via DCR (Dynamic Client Registration):
Run the following curl command in your terminal, replacing <your-redirect-uri> with your client's OAuth redirect URI:
curl -X POST https://slite.com/api/mcp/oauth/register \
  -H "Content-Type: application/json" \
  -d '{
    "client_name": "Gemini Enterprise",
    "redirect_uris": ["<your-redirect-uri>"],
    "response_types": ["code"],
    "grant_types": ["authorization_code", "refresh_token"],
    "token_endpoint_auth_method": "client_secret_post"
  }'
The response will return your client_id and client_secret.
    Configure your client with the endpoint values:
Field
Value
Issuer
https://slite.com/api/mcp/oauth
Authorization Endpoint
https://slite.com/api/mcp/oauth/auth
Token Endpoint
https://slite.com/api/mcp/oauth/token
Registration Endpoint
https://slite.com/api/mcp/oauth/register
Userinfo Endpoint
https://slite.com/api/mcp/oauth/me
JWKS URI
https://slite.com/api/mcp/oauth/jwks
MCP Server URL
https://api.slite.com/mcp
The local (self-hosted) MCP server option has been deprecated and is no longer supported. The remote endpoint https://api.slite.com/mcp is the only supported option.


Supported Tools

Here are the tools you can currently use through Slite MCP. The Slite MCP offers 40 tools for reading and writing to your workspace.
Tool
Description
append-blocks
Add content blocks to existing docs
archive-note
Archive a doc
ask-slite
Get answers from your docs
create-channel
Create a new channel (top-level container for docs)
create-collection
Create a collection database with columns
create-comment-thread
Start a new comment thread on a doc, optionally anchored inline to a specific text span
create-note
Create new docs (with optional parent/channel)
edit-document
Apply multiple block edits to a doc as one atomic change
get-comment-thread-on-note
Retrieve a specific comment thread on a doc, including all replies
get-note
Retrieve full content of a specific doc
get-note-children
List all sub-docs under a parent
get-user
Get user details by ID
get-user-group
Get user group details by ID
list-channels
List all accessible channels in workspace
list-comment-threads
List comment threads on a doc, with filtering by resolved/unresolved state
list-empty-notes-for-knowledge-management
List empty docs for knowledge management cleanup
list-inactive-notes-for-knowledge-management
List inactive docs for knowledge management cleanup
list-notes-for-knowledge-management
List docs with knowledge management filters
list-public-notes-for-knowledge-management
List public docs for knowledge management
list-recently-edited-notes
Show docs you recently edited
list-recently-visited-notes
Show docs you recently viewed
modify-block
Edit a specific block within a doc, in place
modify-range
Replace a range of consecutive sibling blocks within a doc
move-note
Move a doc to a different channel or parent
read-thread
Read an AI thread back as question and answer rounds
remove-blocks
Delete specific blocks from a doc
reply-to-comment-thread
Post a reply to an existing comment thread
resolve-comment-thread
Mark a comment thread as resolved
restore-note
Restore an archived doc
search-notes
Find docs by keyword across your workspace
search-user-groups
Find user groups by name
search-users
Find users by name or email
set-note-review-state
Set the review state and/or review owner of a doc
slite-agent
Run Slite Agent across your workspace knowledge base
unresolve-comment-thread
Reopen a previously resolved comment thread
update-channel
Update a channel's name, icon, or visibility
update-collection
Add or remove columns in a collection database
update-note
Update title or content of existing docs
update-table
Make a structured edit to a table inside a doc
upload-file
Upload a local image or file and embed it in a doc
verify-note
Mark a doc as verified


Uploading files and images

The upload-file tool lets your AI assistant put a local image, plot, screenshot, or PDF straight into a doc. The file is sent directly to Slite storage instead of passing through the AI model, so large files work too.
    Your assistant calls upload-file with the target doc, file name, type, and size.
    Slite returns a short-lived signed upload link, and your assistant sends the file to it.
    The file is embedded as a block in the doc.

Good to know

  • File size follows your plan. Uploads go through the same size check as any other file attached to a doc: 10 MB on Free and Solo, 200 MB on Basic, Standard, and trials, and unlimited on Pro and Enterprise. Images are capped at 20 MB. There is no separate size limit for MCP uploads.
  • The file block can look broken for a moment. It stays broken until the upload finishes — refresh the doc if it still looks that way.
  • The upload link expires. If the upload doesn't complete in time, your assistant has to request a new link and try again.
  • You need edit access to the target doc. The file inherits that doc's permissions, so anyone who can see the doc can see the file.
  • Uploads count toward your workspace's storage.
  • Uploads are rate limited to 20 per minute and 60 per 10 minutes, per user.
  • Files that already have a public URL don't need uploading. Your assistant can embed the URL directly, and Slite saves its own copy.

Managing Tools

You may be able to control which tools your AI assistant can access, depending on your client.

For Claude Desktop

    Go to Settings > Connectors > Slite > Configure , then review when Claude is allowed to use the tools.
  • Always allow — Claude can use this tool without prompting you
  • Needs approval — Claude will ask before using this tool
  • Blocked — Claude cannot use this tool

For other MCP clients

Tool management varies by client. Check your AI assistant's documentation to see if and how you can enable or disable individual tools for a connected MCP server.