# Pome

## Docs

- [Authenticate your API requests to the Pome platform](https://docs.pome.sh/api/authentication.md): Use API keys passed in the X-API-KEY header to authenticate every request to the Pome REST API. Get your key from the team dashboard.
- [Runs API — retrieve scored agent evaluation results](https://docs.pome.sh/api/runs.md): List and retrieve scored Pome run records, including signed artifact download URLs for tool calls, twin state snapshots, and evaluation criteria results.
- [Sessions API — provision and manage twin environments](https://docs.pome.sh/api/sessions.md): Create, list, and delete Pome twin sessions via the REST API. Each session provisions an isolated twin environment with seeded state.
- [Usage API — monitor your team's monthly session quota](https://docs.pome.sh/api/usage.md): Query your team's current session usage and remaining quota with GET /v1/usage to track monthly consumption and avoid hitting plan limits.
- [pome CLI commands: run, inspect, export, and more](https://docs.pome.sh/cli/commands.md): Complete reference for all pome CLI commands — run, inspect, export, init, login, twin, sessions, and version with flags and exit codes.
- [Configure Pome with .pome.json](https://docs.pome.sh/cli/configuration.md): The .pome.json file controls your Pome project's deployment mode, twin connection URL, and default settings used by the pome CLI in every environment.
- [Pome CLI environment variables](https://docs.pome.sh/cli/environment-variables.md): Environment variables consumed by the pome CLI — covering authentication, LLM endpoint configuration, agent-side twin connection details, and debug output.
- [Pome CLI: commands, flags, and installation](https://docs.pome.sh/cli/overview.md): The pome CLI runs agent scenarios, inspects run traces, manages hosted sessions, and authenticates with the Pome control plane from your terminal or CI.
- [How Pome scores and evaluates agent runs](https://docs.pome.sh/concepts/evaluation.md): Pome scores runs locally with your own LLM key — deterministic criteria query twin state directly, probabilistic criteria call your configured LLM judge.
- [Scenarios: define agent tests in Markdown](https://docs.pome.sh/concepts/scenarios.md): A Pome scenario is a Markdown file that defines your agent's task, seed state, and success criteria — the complete specification for one evaluatable run.
- [Sessions: twin lifecycle and isolation](https://docs.pome.sh/concepts/sessions.md): Each Pome run gets an isolated session with fresh twin state seeded from your scenario's Setup block. Sessions expire automatically after their TTL.
- [What is a Pome twin? Resettable API for agents](https://docs.pome.sh/concepts/twins.md): A Pome twin is a stateful, resettable copy of an external API — like GitHub — that your agent tests against without touching production systems.
- [Run Pome scenarios in GitHub Actions](https://docs.pome.sh/guides/ci-integration.md): Add Pome to GitHub Actions using the official run-scenarios-action to run agent scenarios on every push and upload scored trace artifacts automatically.
- [Connect your agent to a Pome twin](https://docs.pome.sh/guides/connect-agent.md): Learn how to point any AI agent at a Pome twin using environment variables, and call the GitHub-shaped REST or MCP surface during a run.
- [Self-host Pome with Docker Compose](https://docs.pome.sh/guides/self-host.md): Run the full Pome twin and CLI locally or on your own infrastructure using Docker Compose — no Pome account or API key required to get started.
- [Write your first Pome scenario](https://docs.pome.sh/guides/write-scenarios.md): Step-by-step guide to writing a Pome scenario file — define seed state, your agent's task, and deterministic or LLM-judged success criteria in Markdown.
- [Install the Pome CLI and authenticate](https://docs.pome.sh/installation.md): Install the pome CLI via npm or bun to run scenarios, inspect run traces, and manage hosted sessions from your terminal or any CI pipeline.
- [Pome: test AI agents against a GitHub-shaped twin](https://docs.pome.sh/introduction.md): Pome gives AI agents a resettable GitHub twin to test against — real REST APIs, 35 MCP tools, deterministic scoring, no rate limits.
- [Pome pricing and licensing frequently asked questions](https://docs.pome.sh/pricing/faq.md): Answers to common questions about Pome's pricing tiers, AGPL license, BYOK LLM evaluation, session quotas, discounts, and migrating between plans.
- [Pome plans, pricing, and tier comparison](https://docs.pome.sh/pricing/overview.md): Pome offers a free OSS self-host tier, hosted Free and Pro plans at $0 and $99/month, and enterprise contracts for compliance-focused teams.
- [Get started with Pome in 60 seconds](https://docs.pome.sh/quickstart.md): Learn how to boot a GitHub-shaped twin with Docker, run the bundled triage agent against it, and inspect your first scored scenario run with Pome.
- [GitHub twin fidelity and known gaps](https://docs.pome.sh/twin/fidelity.md): Understand what the Pome GitHub twin faithfully replicates and what it omits — fidelity tiers, enforced invariants, known deviations, and how to file a gap.
- [GitHub twin: all 35 MCP tools by category](https://docs.pome.sh/twin/mcp-tools.md): The GitHub twin exposes 35 MCP tools your agent can call via the MCP protocol — covering repositories, issues, labels, pull requests, files, and branches.
- [Pome GitHub twin: REST and MCP surface overview](https://docs.pome.sh/twin/overview.md): The Pome GitHub twin is a stateful local copy of GitHub's API — REST routes and 35 MCP tools backed by SQLite, resettable between runs.
- [GitHub twin REST endpoint reference](https://docs.pome.sh/twin/rest-surface.md): The GitHub twin exposes 11 curated REST endpoints that match GitHub's API shape — path, request body, response, and error envelopes.