AgentField
Self-hosted AI agent platform for Python, Go and TypeScript, with cryptographic identities, audit trails and LLM support through LiteLLM. Apache 2.0 licensed.
AgentField is a self-hosted, open source platform for developers building AI agents that need to work together under shared access rules. It runs in your own infrastructure and combines agent execution with identity, policy enforcement and audit records. The project uses the Apache 2.0 license.
Agents can use Python, Go or TypeScript functions, with REST endpoints for other applications to call. AgentField supports LLMs through LiteLLM and returns structured output using Pydantic or Zod schemas. Agents can discover each other's capabilities and make traced calls across the system. Built-in memory handles key-value storage and vector search without requiring Redis.
For long tasks, the platform provides asynchronous execution, streaming, retries and progress updates. Human approval can pause a workflow and resume it later, with state that survives agent restarts. Teams can compare agent versions through A/B testing and gradually shift traffic between deployments. Workflow graphs, execution timelines and Prometheus metrics help trace failures and inspect agent behavior.
Governance is a central feature: each agent has a cryptographic identity, signed requests authenticate calls between agents, and tag-based policies control which agents can call which capabilities. Execution records include tamper-proof receipts that you can verify offline. AgentField also orchestrates multi-turn coding tasks through its own harness or external tools such as Claude Code, Codex and OpenCode.
