HarnessRouter: Unified interface for agent harnesses
Hey HN! We are building HarnessRouter, a canonical API for running Codex, Claude Code, Hermes, and other managed agent harnesses as your product backend. Before building HarnessRouter, I used to build our own agent harness for our products. I tried LangGraph, agent SDKs from different vendors, pydantic, LLM tool use / function call, and so on. It's a very heavy lifting engineering effort, and I am disappointed about the agent deliveries compared to what Codex, CC can deliver. That changed my mindset. The frontier labs and famous open source communities are already putting so much…
In plain words
HarnessRouter is a unified API that lets developers run multiple AI agent systems—including Codex, Claude Code, and Hermes—through a single interface. The self-hosted Community Edition implements the Unified Harness Protocol, an open standard, and handles sessions, streaming, file management, and error handling. It's designed for product teams who want to leverage best-in-class agent harnesses from frontier labs and open-source communities without building their own infrastructure, using their own infrastructure and API keys.
written from the facts on this page · September 2026
From the sources
HarnessRouter Community Edition: the self-hosted, Apache-2.0 edition of the unified interface for agent harnesses. Run Codex, Claude Code, Hermes, PI, DSH, and more through one API, with sessions, streaming, files, cancellation, and failure handling. Implements the Unified Harness Protocol (UHP), an open standard. Your keys, your infrastructure. - HarnessRouter/harnessrouter
In the maker’s words, at launch
Hey HN! We are building HarnessRouter, a canonical API for running Codex, Claude Code, Hermes, and other managed agent harnesses as your product backend. Before building HarnessRouter, I used to build our own agent harness for our products. I tried LangGraph, agent SDKs from different vendors, pydantic, LLM tool use / function call, and so on. It's a very heavy lifting engineering effort, and I am disappointed about the agent deliveries compared to what Codex, CC can deliver. That changed my mindset. The frontier labs and famous open source communities are already putting so much engineering effort to build the world's best harnesses, why not leverage them directly instead of building our own, just like how we call LLM chat completion endpoints instead of training our own models? We provide a docker image to run HarnessRouter locally. ---------- Quickstart: docker pull harnessrouter/harnessrouter docker run -d --name harnessrouter -p 127.0.0.1:3000:3000 -v harnessrouter:/data harnessrouter/harnessrouter docker logs -f harnessrouter Wait for the "ready on :3000" show up, then open the browser at http://localhost:3000. Default username/password is harnessrouter/harnessrouter Then in Integrations page, add your model provider credentials or API keys. In Harnesses tab, as of today we provide routing to Codex, Claude Code, and Hermes as base harnesses. You can customize any of them and configure harness instruction, MCP tools, and skills. Then go to Tasks and let them do jobs. ---------- Every harness has its own request/response format and incompatible with each other. We propose Unified Harness Procotol [1] to standardize how an application talks to an agent harness. It covers harness selection and configuration, task execution, event streaming, sessions start cancel and resume, artifact management and delivery, and failure handling. It's similar idea like LiteLLM, but for harnesses rather than models. HarnessRouter implements UHP. We provide an AGENTS.md [2] and your coding agent can follow it to integrate your application with the harnesses available. We also provide starter kits [3] to demonstrate some types of agentic products that can be built on HarnessRouter. It currently includes PPT agent, Spreadsheet agent, BI Dashboard agent, and Video generation agent. Can't wait to hear what you think! [1] https://unifiedharnessprotocol.org [2] https://harnessrouter.ai/agents.md [3] https://github.com/harnessrouter/starter-kit
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 26d ago · cactuscompute.com
Launched alongside, August 2026
the whole month →- TL
Life & fun · 10d ago · louisabraham.github.io
- SA
Hello HN! I found that picking out plausible but diverse skin tones for my digital art and game development projects was kind of difficult, and I got curious about if there was a way to define a color space that made it easy. I've built a color picker and procedural generation algorithm based on the space as well as a bunch of other fun js features and demos throughout the page that use the equations. If you find it interesting, I have lots of explanations of how I built it and what properties the space has. The methodology might be a bit shaky, but hopefully the result is as helpful for…
Life & fun · Aug 2026 · toneyalexander.github.io
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com