Relato lets you add an external MCP server as an integration. Once it is connected, that service's tools show up in your AI Content Agents, so an agent can read and act in that service the same way it reads and acts in your Relato workspace. The connection flow is the same for every service, so this article is the one place the steps live. Every per-service guide points back here and only covers what is specific to that service.

What an MCP connection gives you

MCP (Model Context Protocol) is an open standard that lets a tool provider expose its actions to an AI agent. When you connect a service's MCP server to Relato, Relato discovers the tools that server offers and makes them available to agents in your organization. From then on an agent can, for example, create a Webflow item, pull an Ahrefs keyword report, or post to a Slack channel as one step in a content workflow.

What you need before you start

An admin role in the Relato organization where you want the integration to live.
The MCP URL for the service you want to connect. It usually looks like https://your-service.example.com/mcp. The service's own docs or admin will have it.
Permission to approve an OAuth connection for that service.
A workspace and at least one agent, or the willingness to create one, so you have somewhere to attach the new tools.

Step 1: Add the service as an MCP account

In Relato, open Settings for your organization, then go to Integrations.
Find MCP in the integrations list and click Add account.
Paste the service's MCP URL.
Approve the OAuth prompt that opens. You will sign in to the service and confirm the connection.
Give the account a clear display name. This name prefixes every tool from that service in the agent tool picker (for example, [Webflow] createItem), so pick something you will recognise later.

Once connected, Relato discovers the tools the service exposes and makes them available to agents in your organization. You can edit the URL or revoke the account from the same Integrations page at any time.

Step 2: Attach the tools to an agent

From the AI Content Agents area, create a new agent or open an existing one. A blank agent is fine.
In the agent's Tools picker, attach the service tools you want it to use. They are listed under the display name you set in step 1.
Also attach the Relato tools the agent reads from, so it has the project context to act on. The query tools and project tools are the usual starting point, plus Google Workspace tools if your drafts live in Google Docs.

Step 3: Set a trigger

Decide when the agent should run. In the agent's Triggers settings you can choose:

Project stage changed, so the agent fires when a project moves through your board (for example, into a "Ready to publish" stage). Set the agent's Scope to the workspace whose projects you want to act on.
A schedule, so the agent runs on a recurring cadence.
Manual, so you run it on demand from the agent thread.

The stage-changed trigger fires for every stage transition in the scoped workspace, not only transitions into one specific stage. Put the "only act when the project reaches stage X" check inside the agent's instructions. That keeps the trigger simple and the routing logic visible in one place.

Step 4: Test with a single project

Pick one project that is safe to act on, or point the agent at a sandbox in the connected service.
Trigger the agent, by moving the project into the trigger stage or running it manually.
Watch the agent's thread. It shows each tool call, including the calls to the connected service, with the payload.
Check the result in the service. Adjust the agent instructions and rerun until a clean pass works end to end.

One caveat: Relato's MCP integration is OAuth-based

Relato connects to MCP servers over OAuth. If a service's MCP server only supports API tokens, you have two options: put an OAuth proxy in front of it, or wait for that service to ship OAuth support. Where a public OAuth MCP does not exist yet for a service you want, tell us at [email protected]. Those requests help us decide which connectors to build or partner on next.

Tips

One account per environment. Add separate MCP accounts for staging and production with different display names, so the agent picker shows [Service Staging] and [Service Production] and you can route workspaces to the right one.
Keep the agent instructions boring. The value is in a precise field mapping and clear rules for missing fields and errors. Spell out what to do in each edge case.
Leave a breadcrumb. Have the agent post a comment on the Relato project with a link to whatever it created, so the team can trace the result.

FAQ

Which services can I connect?

Any service that runs an OAuth-capable MCP server reachable from Relato. See the per-service guides for the ones we have documented, and the OAuth caveat above for the rest.

Where do the connected tools show up?

In the agent tool picker, prefixed with the display name you chose. They are only available to agents you attach them to.

How do I revoke a connection?

Open Integrations, find the MCP account, and choose Revoke. Agents that were using its tools lose access to them immediately.