What Is Deploy & Operations?

Finishing an Ability or Agent on the canvas doesn't automatically connect it to outside systems. To let a backend server or another service actually call your workflow, you need to publish it in a form that's reachable from the outside. Agentria provides this process as a concept called Deploy & Operations.

Deploy & Operations is made up of two branches, API and MCP Server, and covers deploying a finished Ability or Agent so it can be called from the outside, then confirming that the call actually works.

Deploy & Operations: Making a Finished Workflow Callable from the Outside

Under the Deploy & Operations menu in the project sidebar, you'll find two submenus grouped together: API and MCP Server.

Both share the same goal — making a workflow you finished on the canvas callable from the outside — but they differ in the protocol the caller uses. If you want a typical backend server or service to call it directly over REST, expose it as an API. If you want an MCP client like ChatGPT or Claude to call it as a tool, expose it as an MCP Server.

Operational features that come after deployment, like API call logs, aren't a separate menu — they live within the API and MCP Server subpages themselves.

API: Deploying and Calling via REST

Before you can deploy an Ability or Agent as an API, you first need to create a deploy Version. From there, click Publish in the top-right corner of the canvas and select API Release. After setting a name and token, you'll get an API endpoint and API key.

External services authenticate with the issued endpoint using an X-API-KEY header. Abilities are called through the Ability API endpoint, and Agents through the Agent API endpoint. Both support asynchronous execution, SSE streaming, status/result queries, and request cancellation. Unlike Abilities, Agents operate around a Chat Room that preserves conversation history.

If you're new to the API, you can test calls first with a client tool like Postman. Call information is available in the logs on Agentria's API tab.

MCP Server: Exposing to MCP Clients

MCP Server lets you expose a finished Ability or Agent so an MCP (Model Context Protocol) client can call it as a tool. Where the MCP node lets an Agentria workflow connect to an external MCP server as a client, MCP Server does the opposite — it turns your Agentria Ability or Agent itself into an MCP server that external MCP clients can call.

The MCP Server guide is still being written.

Learn More

If you'd like to try deploying and operating a workflow yourself, check out the guides below.

• 🔗Deploy Version — walks through creating the deploy version you need before deploying as an API.

• 🔗API Release Guide — walks through deploying an Ability or Agent as an API.

• 🔗Using the Ability API Endpoint — walks through calling a deployed Ability API from the outside.

• 🔗Using the Agent API Endpoint — walks through calling a deployed Agent API from the outside.

• 🔗API Postman Guide — walks through testing an API call with Postman.

Frequently Asked Questions

What is Deploy & Operations?

Agentria's Deploy & Operations is the process of publishing an Ability or Agent you finished on the canvas so external systems can call it. It offers two paths: API, for calling it directly over REST, and MCP Server, for letting MCP clients call it as a tool.

When should I use Deploy & Operations?

Use it when you want a backend server, another service, or an MCP client like ChatGPT or Claude to call an Ability or Agent you finished on the canvas. Choose API for typical REST integration, or MCP Server if you want it used as a tool by an MCP client.

How is the API Release guide different from the API endpoint guides?

The API Release guide covers publishing an Ability or Agent as an API to get an endpoint and API key. The API endpoint guides cover how to actually call that issued endpoint from an external service. In other words, deployment comes first, and calling the endpoint is the usage step that follows deployment.

How is the MCP Server guide different from the MCP node tutorial?

The MCP node tutorial is a client-side guide where an Agentria workflow connects to an external MCP server to call its tools. The MCP Server guide is the opposite — it covers exposing your Agentria Ability or Agent itself as an MCP server so an external MCP client can call it as a tool. The direction of the call is reversed between the two.