The MCP server exposes Relato as a set of tools the model can call directly: querying projects, tasks, assets, agents, and templates; creating and updating them; running web research; and reading or writing Google Workspace content tied to your projects.
It is designed for content teams who want their AI assistant to operate inside Relato rather than alongside it.
What you can do with the Relato MCP server
Once connected, your assistant can:
Look up projects, tasks, assets, templates, content pillars, and agents in your organization, and filter them by date, status, assignee, or workspace.
Create new projects, tasks, comments, and assets, or update existing ones.
Read and write Google Docs, Sheets, and Slides linked to your projects.
Search the web, run SERP and Reddit queries, crawl pages, and persist findings as Relato assets.
Manage your AI Content Agents: create them, change their scopes, edit their memory files, attach tools, set schedules and triggers.
Read and write Relato datasets and widgets.
Run any Relato LLM query (the same set used by your AI Content Agents internally).
The full tool catalog is large. As of the latest release the server exposes more than 200 tools across queries, commands, integrations, and utilities.
Connecting the server
The MCP endpoint is:
The exact configuration step depends on your client.
Claude (desktop, web, and code): Open Settings, go to Connectors or Integrations, choose Add custom connector, paste the URL above, then approve the OAuth prompt that opens in your browser. You will be asked to sign in to Relato and confirm the connection.
Cursor, Windsurf, Goose, and other MCP clients: Add a remote MCP server entry pointing at the URL above. The client will discover the authorization endpoints automatically and prompt you to sign in on first use.
You only need to sign in once per client. The connection refreshes itself in the background.
How authentication and permissions work
The server uses OAuth 2.1 with PKCE. When you sign in:
You are authenticated as your Relato user.
The access token is scoped to the organization you are signed in to.
Every tool call is org-isolated. The assistant can only see and modify data belonging to that organization, even when it asks for something by ID.
Tokens expire after an hour and refresh transparently. You can revoke a connection at any time from your Relato account settings.
The assistant cannot see your password, your Google credentials, or data from other organizations you may also belong to.
Tool categories
The tools the model can call are grouped into a small number of categories. You do not need to memorise them, but it helps to know what is in scope when you ask the assistant for help.
Category | What it covers |
Queries | Read-only lookups across projects, tasks, assets, agents, templates, organizations, members, workspace folders, agent memory, and agent conversation starters. |
Project and task management | Create, update, assign, and move projects and tasks. Set due dates, stages, statuses, content types, and labels. |
Workspace and templates | Create workspaces and folders, manage stages, duplicate templates, manage template tasks. |
Assets | Create assets from URLs, Google Docs, and notes. Link assets to projects, tasks, and workspaces. Manage labels. |
Agents | Create and clone agents, edit their prompts, memory files, scopes, conversation starters, schedules, and triggers. |
Comments and threads | Read agent thread messages, post comments, react, request user input. |
Web research | Web search, SERP, AI Overview lookups, Reddit search, crawl, extract, fetch. |
Google Workspace | Read and write Docs, Sheets, Slides, and Drive files and comments connected to the user's Google account. |
Datasets and widgets | Create, query, and update structured datasets and dashboard widgets. |
Tips for working with the server
Tell the assistant which workspace or project you are working in. It speeds up the right tool calls and avoids ambiguity.
For bulk imports or migrations, write a short instruction document describing your column-to-field mapping and paste it as context. The assistant will use it as the source of truth.
Read-only queries are free of side effects, so it is safe to ask the assistant to explore and summarise before you authorise any writes.
Long-running operations such as scans and crawls happen in the background. The assistant will get back a result asynchronously.
FAQ
Which AI clients does it work with? Any client that supports the Model Context Protocol over HTTP. That includes Claude (desktop, web, and code), Cursor, Windsurf, Goose, and the MCP Inspector.
Does the assistant see data from other organizations I belong to? No. Each access token is scoped to a single organization. To work in another organization, sign in to that one in Relato first and reconnect from your AI client.
Can I limit what the assistant can do? The token honours the same permissions as your Relato user. If you do not have write access to a workspace, neither does the assistant. You can also revoke the connection from your Relato account settings at any time.
Why are some tool calls slow? Tools that hit external services, such as web crawls, SERP lookups, and Google Workspace writes, can take several seconds. Most Relato queries return in well under a second.
How do I disconnect? Open your Relato account settings and revoke the MCP connection. You can also remove the connector from your AI client directly.
Related articles
If you need help, message us in the Relato Slack community or email [email protected].