# Agent Jig

> Evaluation and testing framework for AI agents. Define test scenarios in YAML, run against your agent in CI, get pass/fail scores and regression diffs.

Agent Jig is a developer tool that solves the "silent regression" problem in AI agent development. Teams can write eval scenarios in YAML, run them against any agent (LangChain, LlamaIndex, CrewAI, custom Python/Node, or any HTTP endpoint) via a CLI, and integrate results into CI pipelines for automatic regression detection.

## Core Concept

A "jig" is a tool that holds a workpiece still so every operation is repeatable. Agent Jig holds your AI agent still during testing — same inputs, deterministic scoring, locked baseline — so you know before every deploy whether your agent got better or worse.

## Key Features

- YAML-defined eval scenarios (human-readable, version-controllable)
- CLI runner: `agent-jig run --config eval.yaml`
- Deterministic pass/fail scoring per scenario
- Baseline locking: lock a known-good run, diff all future runs against it
- CI integration with GitHub Actions and CircleCI
- Regression alerts via Slack and email (Pro)
- Framework-agnostic: works with any agent that accepts input and returns output
- Score history and trend charts (Pro)

## Pricing

- Free: 50 evals/month, up to 10 scenarios, CLI runner
- Pro: $89/month, 5,000 evals, unlimited scenarios, CI integration, regression diffs, baseline locking
- Enterprise: Custom pricing, unlimited evals, SSO, private cloud option

## Pages

- [Home](https://agentjig.com/)
- [Pricing](https://agentjig.com/pricing.html)
- [Sign Up](https://agentjig.com/signup.html)
- [About](https://agentjig.com/about.html)
- [Blog](https://agentjig.com/blog/)
- [Blog: AI agent evaluation 101](https://agentjig.com/blog/ai-agent-evaluation-101.html)
- [Blog: How to write eval cases that catch regressions](https://agentjig.com/blog/eval-cases-that-catch-regressions.html)
- [Blog: The silent regression problem](https://agentjig.com/blog/silent-regression-problem.html)
- [Blog: Integrating evals into your CI pipeline](https://agentjig.com/blog/integrating-evals-into-ci.html)
- [Blog: What a 95% pass rate actually means](https://agentjig.com/blog/what-95-percent-pass-rate-means.html)
- [Privacy Policy](https://agentjig.com/privacy.html)
- [Terms of Service](https://agentjig.com/terms.html)

## Contact

- press@agentjig.com (press inquiries)
- jobs@agentjig.com (careers)
- enterprise@agentjig.com (enterprise sales)