HarnessRouter open-sources Harness Arena: benchmark one task across multiple agent harnesses and models side by side
blaizedsouza · x · 2026-11-04
The HarnessRouter team released Harness Arena, letting you test the same task across multiple harness × model combinations side by side.
- The pitch: we benchmark models constantly, but the harness running the model dramatically affects results, cost, and execution time.
- How it works: every config gets the same prompt, files, tools, skills, and evaluation checks; the arena then reports total cost, execution time, pass/fail, and produced files and artifacts.
- It's built on HarnessRouter (Apache-2.0, 4.1k GitHub stars), a plug-and-play unified interface that runs Codex, Claude Code, Hermes, PI, DSH and more through one API via the open Unified Harness Protocol (UHP), with sessions, streaming, files, cancellation and failure handling — your keys, your infrastructure.
More from coding & agent
- Living Brain tracks 3 health scores to stop AI agents relying on stale memory — cneuralnetwork · 2026-11-04
- Open-source skill makes Claude Code ship polished, notarized macOS apps in one shot — vista8 · 2026-11-04
- Power user tallies 12B coding-agent tokens, finds 96% were cache reads — peakpirate007 · 2026-11-04
- Indian code-mixed calls have no single correct transcript: LLM, CRM and compliance each need a different one — BEYOND__7 · 2026-11-04
- Alexey Grigorev releases free kit of 28 prompts to ship an app from spec to production — Al_Grigor · 2026-11-04
- Hugging Face Hub MCP Server lets agents search models and datasets via MCP — modelcontextprotocol · 2026-11-04