Get startedRequest a demo

Get started

From zero to a first evidence bundle.

Two ways in. Hand it to the coding agent you already use, or run four commands yourself. Either way you end the afternoon with a decision and a bundle you can show someone.

With a coding agentBy handRunner docs

An OpenAI-compatible endpoint

vLLM, TGI, Ollama, a gateway: anything that speaks the API, reachable from the machine you run on.

Python 3.11+ and uv or pip

The runner is a Python CLI. No containers, no database, no service to stand up.

No account, no key from us

The runner is Apache-2.0 and never calls out. Nothing to register, nothing to phone home.

Option 1

Hand it to your coding agent.

Cursor, Claude Code, Codex or whatever you run: paste the prompt, or point the agent at the skill file. The skill tells it how to install the runner, reach your endpoint, run a pack, verify the bundle and report the decision without sending anything anywhere.

Open skill.md
Prompt to paste
Install the Etalon runner from https://github.com/coyos-ai/etalon following its README.
Run the example pack "contact-routing" from the repo's examples/ directory against my OpenAI-compatible endpoint at https://llm.internal/v1 and write the bundle to ./evidence.
Run etalon verify on the bundle.
Summarise decision.json, coverage.json and fingerprint.json for me.
Do not send anything to third-party services.
Or install the skill
# into your agent's skills directory, for example:
mkdir -p .cursor/skills/etalon
curl -fsSL https://etalon.coyos.ai/skill.md -o .cursor/skills/etalon/SKILL.md

The skill defers to the runner README for exact flags, so it stays correct as the CLI evolves. Review it before use, like anything else an agent will execute.

Option 2

Four commands, by hand.

The example pack ships in the repository so you can see a full run before licensing anything. Enterprise packs are directories delivered to you; they run the same way.

Commands mirror the runner README. If they ever disagree, the README wins.

  1. 1

    Install the runner

    git clone https://github.com/coyos-ai/etalon
cd etalon
uv sync   # or: pip install -e .
  2. 2

    Point it at your endpoint

    Any OpenAI-compatible URL reachable from this machine. If the endpoint needs a key, export it as described in the README.

    export ETALON_ENDPOINT=https://llm.internal/v1
  3. 3

    Run the example pack

    uv run etalon run --pack examples/contact-routing --endpoint $ETALON_ENDPOINT --out ./evidence
  4. 4

    Verify and read

    Copy ./evidence anywhere, including a machine with no network, and verify it there. Open report.html in any browser.

    uv run etalon verify ./evidence
open evidence/report.html

Then

Swap in the packs your reviewers will ask for.

Want us on the call for the first run?

Thirty minutes with an engineer. We run a pack against your endpoint together and open the bundle.

Request a demoGet started