Agents
Sume developer tools use one boundary: the public API is the source of truth, and wrappers stay thin.
Tooling layers
- Developer API: workspace-scoped HTTP endpoints.
- CLI: shell-friendly wrapper around the Developer API.
- Hosted MCP: remote MCP at
https://mcp.sume.com/mcpfor Cursor, Claude, and similar clients. - Local
sume mcp: future CLI surface (coming_soonin current releases) — use hosted MCP or direct CLI commands instead. - Dashboard tools: app-integrated tools with server-resolved workspace context.
Agents should use catalog, jobs, media schema, and usage reads for planning before requesting confirmation for write or paid actions. See Safe automation.
Where to start
| Goal | Start here |
|---|---|
| Call a saved authoring recipe from your backend | Format API · Format catalog |
| Interactive Cursor / Claude connector | MCP quickstart |
| Local shell agent with CLI | Quick start · CLI overview |
| Backend / server integration | Quick start (API) · Public API |
| Run the Agent on an ad-hoc task from your backend | Agent Completions |
| Run an agent on a recurring schedule | Scheduled |
| Create an API key | API Keys dashboard |
| Agents product UI | Agents |
| Avatar or Image first job | Create new avatar · Image 1.0 |
Hosted MCP paid tools today are still largely Avatar-focused. For Image, Video, or Music generation from an agent, call the Developer API (or a thin tool you own over it). Details: MCP overview.
Workspace context
Agent tools should not ask the user for a workspace id. Workspace context comes from the API key or authenticated app session.