OpenAPI 3.1
All 110 API routes are covered by scoped tokens: owner, operator and read.
Surfaces · API, MCP and CLI
Built for agents to operate, too.
Everything you can do in the console you can do from code, a script or another AI agent, without screenshots.
Live in productionTimeControl
AgentOver MCP: which tasks are blocked in project Idp?example
illustrative
What is API, MCP and CLI in agnte?
agnte is AI-operable. It exposes an OpenAPI 3.1 API with 110 routes covered by scoped tokens, an MCP server with 26 tools, an llms.txt file and a CLI with JSON output, so scripts and other AI agents can run agnte directly.
Capabilities
What API, MCP and CLI does
TimeControl
MCP server, 26 tools
Other agents can plan, query and control agnte through the Model Context Protocol.
CLI with JSON output
Script anything and pipe the results into your own tooling.
Safe by default
Destructive actions need a confirm or a dry run, with idempotency keys and rate limits.
Proof from production
Running today, measured.
110API routes with scoped tokens
26MCP tools
16/16AI-operability golden checks passed
Source: agnte release notes and golden scenario G22.
How it works
How API, MCP and CLI works
- Create a scoped token
- Call the API, MCP or CLI
- Confirm destructive actions
- Read the journal
FAQ
Questions about API, MCP and CLI
Can another AI agent run agnte?
Yes. The MCP server, llms.txt and JSON CLI are designed so agents can operate agnte without screenshots.
How are API tokens scoped?
Tokens are owner, operator or read. Destructive calls need confirmation or a dry run.
Is there an OpenAPI spec?
Yes, OpenAPI 3.1 covering all 110 routes.
Related