AgentField

AgentField is a self-hosted AI agent platform for developers building services that need to coordinate tasks under shared access rules. It combines execution, identity and audit records in infrastructure you control.

• Agent services: Build agents with Python, Go or TypeScript functions exposed as REST endpoints, and let them discover and call each other's capabilities with tracing.
• Model access and memory: Use LLMs through LiteLLM, return structured Pydantic or Zod output, and retain data with built-in key-value storage and vector search.
• Durable workflows: Run long tasks with streaming, retries and progress updates, or pause for human approval with state that survives agent restarts.
• Identity and access: Give each agent a cryptographic identity, authenticate calls with signed requests and enforce tag-based access policies; execution receipts can be verified offline.
• Deployment oversight: Compare agent versions through A/B testing and gradual traffic shifts, and inspect workflow graphs, execution timelines and Prometheus metrics.

AgentField is open source under Apache 2.0 and supports private deployment. It can also orchestrate multi-turn coding tasks through its AForge harness or external tools such as Claude Code, Codex and OpenCode. LLM access depends on the selected provider; offline verification applies to audit receipts.