BeeAI Framework

BeeAI Framework is for developers building AI agents and systems where several agents work together. It lets developers enforce rules on agent behavior while retaining the model's ability to reason.

• Agent constraints: Set requirements agents must follow across different LLM backends, rather than relying only on instructions in a prompt.
• Multi-agent workflows: Coordinate agents with parallel work, retries and replanning, or describe orchestration in YAML.
• Tools and document retrieval: Give agents built-in or custom tools for actions such as web search and code execution, and use vector stores and document processing for retrieval-augmented generation.
• MCP and A2A: Equip agents with MCP tools, build compatible components and connect with other MCP or A2A agent systems.
• Monitoring: Inspect agent behavior through events, logs and OpenTelemetry traces for monitoring and auditing.

The Python and TypeScript implementations have feature parity. Conversation memory, caching and resource management support applications that host agents on servers.

BeeAI is open source under Apache 2.0 and is hosted by the Linux Foundation under open governance. Ollama supports local model inference; OpenAI, Groq and Watsonx.ai provide cloud backends. Your choice of backend determines where model requests run.