HarnessRouter Community Edition
An open-source agent infrastructure for self-hosting multiple AI harnesses through a unified API and Docker-based deployment.
An open-source agent infrastructure for self-hosting multiple AI harnesses through a unified API and Docker-based deployment.
What the product does and how it is positioned
HarnessRouter Community Edition is an open-source agent harness infrastructure designed for self-hosting on private environments. It allows developers to plug various AI models and agents into their products through a single, unified API.
The system is built around the Unified Harness Protocol (UHP), which standardizes how agent backends interact with different harnesses. By using a containerized approach, it is designed so that all data, including API keys and session history, stays within the user's own infrastructure.
Source-supported ways to use the product
Building applications that require project-aware tasks, streamed output, and persistent sessions for follow-up work.
Creating product experiences that utilize conversational interfaces with built-in templates and validation skills.
The documented workflow, where available
Pull the official Docker image and run the container to start the Gateway, Runner, and Console simultaneously.
Log in to the self-hosted API using configured credentials to establish a secure session.
Connect to a specific harness like Codex or Claude Code using the unified API contract.
Submit inputs and metadata to initiate agent tasks, following the progress via streamed tokens and events.
The Unified Harness Protocol (UHP) is the core specification that enables portability across different agent implementations. It provides a stable, reviewable HTTP contract that is intended to reduce vendor lock-in by standardizing how requests and responses are structured.
By utilizing OpenAPI 3.1 and JSON Schema 2020-12, UHP offers machine-readable definitions that allow for strong validation of every object. This protocol is supported by 52 runnable checks that developers can use to verify conformance locally.
Checks to run with your own material and workflow
What was checked and when
Answers based on the source-checked product record
It is an open, versioned HTTP contract based on OpenAPI 3.1 that makes agent harness implementations portable and testable.
It is deployed as a single Docker container that includes the API Gateway, Runner, and management Console.
The platform supports integration with harnesses such as Codex by OpenAI, Claude Code by Anthropic, and Hermes by Nous Research.
All API keys, session states, and workspace files are stored locally on the user's infrastructure using SQLite and a workspace volume.
Yes, the system supports streaming tokens and events throughout the agent's work cycle to provide immediate feedback.