Introduction

Setu is the layer between AI agents and the apps people use. Your agent calls github.create_issue or slack.post_message; Setu handles the sign-in, keeps the tokens safe, and makes the API call.

What it does#

• Sign-in. Setu runs the OAuth flow for each app, including state, PKCE, callbacks and token refresh. Your user clicks Authorize once.
• Token vault. Tokens are encrypted at rest, refreshed before they expire, and erased when a user disconnects.
• Tools. Each app is a set of actions with validated input and typed errors. Search the catalog, hand the results to a model in OpenAI or Anthropic format, and run what it calls through one endpoint.
• MCP. The same tools are available to Cursor, Claude Desktop and other clients over MCP.

What it is not (yet)#

Setu is pre-1.0. Be aware of three things before you build on it:

• The eight launch integrations have been tested against fakes, not against every live provider edge case. Expect fixes.
• The hosted service at api.setuagent.com is rolling out. Until it is open to everyone, access is by invitation: write to [email protected].
• There has been no third-party security audit. How tokens are protected lists what is and is not done.

Where to go#