Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.pome.sh/llms.txt

Use this file to discover all available pages before exploring further.

Pome is a digital-twin testing platform for AI agents. Run your agent against a local GitHub-shaped twin with real REST routes and 35 MCP tools, backed by SQLite state you can reset between every run. No GitHub rate limits, no shared sandbox, no flaky network — just deterministic, repeatable evaluations.

Quick Start

Boot your first twin and run an agent against it in under 60 seconds.

Write Scenarios

Define your agent’s task, seed state, and pass/fail criteria in a single Markdown file.

CLI Reference

Full reference for all pome commands — run, inspect, export, and more.

API Reference

Authenticate and interact with the Pome REST API to manage sessions and runs.

​
How Pome works

1

Boot a twin

Start a GitHub-shaped twin locally with docker compose up or provision one via the hosted API. The twin exposes real REST endpoints and MCP tools backed by SQLite.
2

Write a scenario

Describe your agent’s task, seed initial state (repos, issues, labels), and define success criteria — all in a single Markdown file.
3

Run your agent

Point your agent at the twin’s URL. Pome records every tool call and API request automatically.
4

Score and inspect

Pome evaluates the final twin state against your criteria and writes a score, trace, and state snapshot you can inspect or export.

Core Concepts

Understand twins, scenarios, sessions, and evaluation before you build.

CI Integration

Drop Pome into GitHub Actions with one step using the official action.

Self-Host

Run the full Pome stack locally with Docker Compose — no account required.

Pricing

Free self-host tier, hosted free tier, and Pro plans starting at $99/mo.