agentfield vs harnessrouter
Open-source control plane that runs AI agents as microservices: write plain Python/Go/TS functions, get REST endpoints with routing, queues, retries, memory and tracing — one laptop to 10k agents. — versus — Self-hosted Apache-2.0 reference implementation of the Unified Harness Protocol: one OpenAI-Responses-style API running Codex, Claude Code, Hermes, Pi and DSH with sessions, streaming, files, traces.
Both are control planes exposing agents as services with REST endpoints; agentfield wraps your own functions, harnessrouter wraps whole coding harnesses.
| agentfield | harnessrouter | |
|---|---|---|
| Stars | 2.6k | 1.1k |
| Forks | 413 | 112 |
| Language | Go | Python |
| License | Apache-2.0 | Apache-2.0 |
| Last activity | 2 days ago | yesterday |
| Topics | agents, orchestration | orchestration, agents, gateway |
| Curated connections | 10 | 5 |
agentfield — the curator's take
The 'agents as a backend' play: write plain functions (no DSL, no graph wiring), and the Go control plane turns each into a REST endpoint any service can call — with fan-out to thousands of parallel agents, queues, retries, versioned deploys, observability and identity/audit built in. Reach for it when agents must be production infrastructure callable by frontends, cron jobs and other services — not a chat window. NOT for notebook experiments or a single local agent (a control plane + SDK is real operational commitment), and it won't give you reasoning-pattern libraries — you still design the agent logic it hosts. Its prompt-to-backend flow (/agentfield in Claude Code/Cursor) is a nice on-ramp, but evaluate the runtime, not the demo.
harnessrouter — the curator's take
Use it when your *product* needs to run coding-agent harnesses as a backend service and you don't want an integration per harness: one Docker image ships Console + Gateway + Runner, installs the harness CLIs you enable, and exposes runs/sessions/artifacts through UHP so you can swap Codex for Claude Code (or a cheaper harness×model pair — their benchmarks show 90%+ cost spread) without touching your app. Bring your own provider keys. It is a server for embedding harnesses, not a desktop cockpit for a developer (see agent-orchestrator, helmor) and not a model gateway (put litellm/omniroute underneath). Community Edition of a commercial product; check what stays in the paid tier before betting on it.