Skip to main content
Cadenya is a headless agent runtime for building and managing AI agents that use tools to accomplish tasks. You define what an agent can do, how it should behave, then let it execute work on your behalf.

​
Core concepts

Cadenya is built around three core concepts:
  • Tool Sets. Collections of tools that give agents capabilities. Point one at an MCP server or OpenAPI spec and Cadenya syncs the tools it finds, or wrap a plain HTTP API by defining each tool yourself.
  • Agents. How an AI executes work. Each agent has one or more variations: configurations with a system prompt, a model, and assigned tools.
  • Objectives. A single run where an agent works on a task. You provide an initial message, and the agent reasons through the problem, calling tools as it needs them.
These concepts stack: you create tool sets to define capabilities, wire them into agents that define behavior, then run objectives to execute work.

​
Get started

Follow the Connect an MCP server lesson to create your first tool set, then Build an agent that improves to configure an agent and run an objective. Or dive into the individual guides:
  • Tool sets: how to organize and manage tools
  • Agents: how to define agent behavior
  • Objectives: how to run and monitor agent work

​
API reference

The API reference documents every endpoint, generated from the Cadenya API specification. Each one ships copy-paste snippets in TypeScript, Go, Ruby, and cURL.