What Tablif read on OpenHarnX's site
OpenHarnX describes itself as: “OpenHarnX: check what your coding agent actually passed”
- Stage: Live, no waitlist
- How it differs: “No model grades its own work.”
What it is
Calls itself “verifier”.
- Calls itself: verifier
“An open-source verifier for changes written by coding agents.”
Where it puts itself
Its one line of difference: “No model grades its own work.”
“No model grades its own work.”
What it costs
No free plan.
- Free plan: no
“No model calls”
Who it sells to
Sells to developers. Available in 4 languages.
- Sells to: developers
- Languages: 4 languages
“Works with any agent via CLI or CI · Claude Code hook built in”
“Experimental TypeScript, JavaScript, Go”
How it runs
A command-line tool for Linux and macOS. Runs locally and works offline. Open source. Says it uses no AI.
- Kind: cli
- Runs on: Linux · macOS
- Runs where: on the device, offline
- Open source: yes
- AI role: explicitly no AI
“Works with any agent via CLI or CI · Claude Code hook built in”
“…macOS arm64 locally · Linux in CI”
What it works with
Plugin for Claude. Names Codex, Cursor and 4 more on its pages.
- Plugin for: Claude
- Names: Codex · Cursor · GitHub · +3
“Claude”
“Codex”
What it offers as proof
Live, no waitlist. Publishes docs.
“No model calls”
“Docs -> https://openharnx.dev/docs”
Sources
Tablif read these pages, last on 8 Oct 2026. Everything above is what the product says about itself; Tablif does not verify every claim.
Similar products in Tablif Town
- DidWork: Verify agent outcomes, not just tool calls.
- DemoKit: Open-source demo videos by your agent — verified, or no file
- TestMyVibe: Audit your product flows with MCP and catch issues early
- GitHub: The AI analyst that shows its work.
- OpenTrustBench: A trust score for every AI agent
- Azdaja: Give your coding agent a persistent Python workspace
- CodeCrab: 100% local-first AI pull request reviewer
- OpenDecider: Open decision models that route AI agents without LLM calls