Most agent frameworks are monoliths.
Jido is composable.
Use the full stack or pick the packages you need. Foundation packages for LLM handling, core framework for autonomy, and specialized packages for AI and coding workflows.
┌─────────────────┐
│ jido_coder │ ◀── AI coding workflows
│ file ops, git │
└────────┬────────┘
│
┌────────▼────────┐
│ jido_ai │ ◀── LLM-powered agents
│ token tracking │
└────────┬────────┘
┌─────────────────┼─────────────────┐
│ │ │
┌──────▼──────┐ ┌──────▼──────┐ ┌──────▼──────┐
│ jido │ │jido_action │ │jido_signal │
│ bot core │ │ validation │ │ pub/sub │
└──────┬──────┘ └─────────────┘ └─────────────┘
│
┌────────────┴────────────┐
│ │
┌──────▼──────┐ ┌───────▼──────┐
│ req_llm │ │ llmdb │
│ HTTP client│ │ model registry│
└─────────────┘ └──────────────┘
Model registry and metadata. Token limits, pricing, capabilities for OpenAI, Anthropic, Google, and custom providers.
HTTP client for LLM APIs. Built on Req with automatic retries, rate limiting, streaming, and function calling.
BEAM-native bot framework. OTP supervision, isolated processes, message-passing. Run 10k+ agents per node.
Schema-based action validation. Required fields, defaults, type constraints. Composable action primitives.
Pub/sub signaling between agents. Decoupled coordination via BEAM message-passing. No external broker.
LLM-powered agents with token/cost tracking, tool calling, and streaming. Integrates jido + req_llm + llmdb.
Ready to build?
Start with the getting started guide or explore production examples.