

Plug-and-play managed agent harnesses in your product: Codex, Claude Code and Hermes through one API on your infrastructure, now open sourced under Apache 2.0. Switch harnesses without rebuilding your backend. Sessions, streaming, files, artifacts, cancellation, failure recovery, all handled. Model provider keys, state, and deliveries under your control. Gateway, Runner, Console in one Docker container. Ready-to-use starter kit apps as inspiration of possibilities.
Loading comments…
Project Info
Product Keywords
HarnessRouter Community Edition is an open-source agent harness infrastructure that lets you plug Codex, Claude Code, and Hermes into your product through a single Agent API—all running on your own infrastructure. Released under Apache 2.0, it packages the Gateway, Runner, and Console into one Docker container, so you can go from image to running in minutes without a managed database, external vault, or cloud account. The API, Console, SQLite state, and workspace files stay on the infrastructure you control, keeping your model provider keys, state, and deliveries firmly in your hands.
Integrate once, then select the harness for each task—Codex by OpenAI, Claude Code by Anthropic, or Hermes by Nous Research—and switch without touching your product backend. The shared contract abstracts away harness-specific differences, so your integration stays stable even as underlying tools evolve.
HarnessRouter handles the backend lifecycle around each harness, not just the initial model request. Sessions can be created, resumed, and inspected; tokens and events stream in real time; files upload and persist in workspaces; artifacts return as finished outputs; cancellation stops running work through one contract; and structured errors are handled consistently across all harnesses.
UHP is the public, versioned contract that makes harness implementations portable, inspectable, and independently testable. It includes a stable HTTP spec, OpenAPI 3.1 machine-readable definitions, JSON Schema 2020-12 contracts for every object, and 52 runnable checks you can verify locally—no lock-in, no guesswork.
Instead of starting from an empty repo, explore complete reference demos and production kits. The Cursor-style coding app shows project-aware tasks, streamed output, persistent sessions, follow-up work, cancellation, and generated-file downloads. LumaCare applies the same architecture to a family-care product, and the Slides Kit offers a conversational copilot with templates and deck validation—all launchable from the HarnessRouter Console.
One container, one API, your infrastructure—agent harnesses without the rebuild.
That's the core promise: HarnessRouter collapses the complexity of running multiple agent harnesses into a single Docker container and a single API contract. You get production-grade lifecycle handling—sessions, streaming, files, artifacts, cancellation, failure recovery—without stitching together disparate tools. And because it's Apache 2.0 with a public protocol, you're never locked into a vendor's roadmap; you can verify conformance locally and move freely between harnesses as your needs change.
You're building a product that needs agent capabilities and want to avoid re-architecting every time a new harness emerges. If you value self-hosting, data control, and open standards, HarnessRouter Community Edition gives you a production-ready foundation with starter apps to accelerate your first implementation.
Other tools you might consider
Ditch your scraper. One API gives your code everything it needs from the web: structured JSON, clean markdown, cited research, and browser automation. No browser, LLM, or pipeline for you to run. Use it from the tools you already work in: an MCP server, CLI, Raycast extension, or as an Agent Skill. Grab a key and make your first call in less than three minutes. Mozilla-backed. Your data is never sold, never trained on.
Navegación basada en árboles, refactorización semántica, análisis de código base y herramientas de portabilidad de lenguajes para agentes de codificación que admiten 163 gramáticas.
GitHits gives coding agents access to the open-source code your app depends on. Get real implementation examples, dependency source navigation, package inspection and documentation. Agents can grep and read your codebase. They can't grep and read the open-source code your app depends on. That's where they start guessing, retrying, and looping. GitHits builds a version-aware index on demand. Agents can search, navigate, and inspect the code behind their dependencies. CLI: npx githits@latest init
Octopoda is an open source infrastructure layer that gives AI agents persistent memory, automatic loop detection, and full observability. Agents forget everything between sessions. Octopoda fixes that with a remember/recall API that survives restarts, crashes, and deployments. The loop detection system monitors 5 signals to catch agents stuck repeating themselves before they burn your API budget, with real-time cost estimation showing exactly how much each loop wastes. A built-in dashboard shows every agent's health score, memory explorer with version history, audit trail logging every decision with reasoning, and a timeline replay that lets you scrub through everything your agent did step by step. Works with LangChain, CrewAI, AutoGen, and OpenAI Agents SDK with one-line integrations. Runs locally with SQLite or connects to cloud with one environment variable. Free tier, MIT licensed. 177 signups, 120 GitHub stars, zero marketing spend.
Maker
indie_inkwell
Loading comments…