Stedia in Claude & Cursor

Stedia is an MCP server. Point Claude, Claude Code, Cursor, or any other MCP client at it, and that agent can work with your Stedia workspace without leaving the window it is already in.

This is the opposite direction from MCP servers, where you add somebody else's server to Stedia.

ConnectCopy Anchor Link

The server address is https://api.stedia.ai/mcp.

Signing in happens in your browser, once per client. Your agent never sees a password. A client that goes unused for 30 days loses its access, and so does every client of a person who leaves the workspace.

After you sign in, Stedia shows which app is asking and where the access goes. The screen lists what the app may do: Read your chats, schedules and assistants, Start tasks with your tools, and Approve actions waiting on you. Nothing connects until you click Allow.

Everything the agent does runs as you, in your workspace.

What your agent can doCopy Anchor Link

Answer questions about Stedia. It searches and reads this help center, so it can explain a feature without guessing.

Read your chats. It can list your Stedia chats and read one in full, with list_threads and read_thread. Long chats are cut short. Only connect an app you would trust with your chat history.

Work with your schedules. List them, read one in full, create one, test it, put it live, pause it, delete it, and read the history of what each run did.

Hand over a task. ask_stedia gives Stedia a job in plain language and returns straight away with a task id. Stedia runs it against your connected tools, and your agent polls get_answer until it is done. Nothing blocks, so a task that takes minutes is fine.

Bring you an approval. When a run stops on an action that needs your decision, get_answer returns the call and its details. Your agent shows you what would happen, you answer, and it passes your answer back with approve_task.

See the workspace. Which tools you have connected, which knowledge bases exist, and which assistants your workspace has published.

Draft an assistant. It can write the assistant's instructions and save them as a draft. Publishing stays in the web app.

A new schedule does not run yetCopy Anchor Link

A schedule your agent creates is a draft. It only starts running once a test run has succeeded and somebody activates it. That is the same rule the web app follows: nothing runs unattended under instructions no test has proven.

Your agent can do both steps. Ask it to test the schedule, then to activate it once the test comes back clean.

ApprovalsCopy Anchor Link

Stedia stops before an action that reaches outside it, like sending mail or writing to a repository, and waits for a person. That gate holds over MCP too.

What changes is where you answer. get_answer hands your agent the waiting call together with its arguments, so the agent can tell you what would actually happen. You decide, and the agent records your answer with approve_task. Denying abandons that one call, and the run carries on.

The decision is yours, not the agent's. Tool descriptions say so, but a model can still be talked into deciding for you, so treat an approval that arrives without a clear question as one to check in the web app. Every approval is also still on the conversation page, and whichever side answers first wins: a call cannot be decided twice.

What still needs the web appCopy Anchor Link

Publishing an assistant. A drafted assistant goes live only after the checks that run in the builder.

TipsCopy Anchor Link

Try these in Claude or Cursor once Stedia is connected:

• "Ask Stedia which of my Linear issues have no assignee, and show me the list."
• "Create a Stedia schedule that summarizes my unread Gmail every weekday at 7:30. Test it and show me the result before you activate it."
• "How do I add a knowledge base in Stedia?" Your agent looks it up in this help center.

When your agent brings you an approval, read what the action would do before you answer. If you are unsure, open the chat in the web app and decide there.

If something goes wrongCopy Anchor Link

See alsoCopy Anchor Link

• Scheduler, for what a schedule is and how the test gate works.
• MCP servers, for adding someone else's server to Stedia.