# XORCISE

> Evaluate cyber AI agents against realistic mission environments — locally on your machine.

## Docs

- [Agents](https://docs.xorcise.ai/concepts/agents.md): An agent in XORCISE is a named record, and the harness recorded on it selects how XORCISE replays its telemetry, correlates it, and launches it.
- [How grading works](https://docs.xorcise.ai/concepts/grading.md): Every XORCISE run scores out of two equal halves — deterministic checks and an LLM judge — and with no judge configured the ceiling is 50%.
- [How XORCISE works](https://docs.xorcise.ai/concepts/how-it-works.md): What xorcise up actually starts: one local process with two network planes, the containers behind a run, and where every file lands.
- [Missions](https://docs.xorcise.ai/concepts/missions.md): A mission is an environment plus the rules for grading it. XORCISE ships two types, lab and static, from your own bundle or the free library.
- [Runs](https://docs.xorcise.ai/concepts/runs.md): A run pairs one agent with one mission under a time budget. XORCISE persists only two states, and the terminal trigger records how the run ended.
- [Traces and events](https://docs.xorcise.ai/concepts/traces.md): XORCISE ingests OpenTelemetry from your agent on two signals and replays it as a timeline, a terrain map, and a readable event feed.
- [Architecture](https://docs.xorcise.ai/contribute/architecture.md): The constraints and invariants the XORCISE codebase depends on — the rules to hold, and the reasons obvious-looking simplifications are wrong.
- [Run a mission from a source checkout](https://docs.xorcise.ai/contribute/run-from-source.md): The extra prerequisites, stub mode, and failure modes that apply when you drive the full run loop from a development checkout rather than an installed package.
- [Evaluate Agents with Playbooks](https://docs.xorcise.ai/experimental/playbooks.md): Run several models through the same missions under one fixed harness, and get a single eval card ranking them — using the xorcise-playbooks skill.
- [Author a mission](https://docs.xorcise.ai/guides/author-a-mission.md): Go from an empty directory to a complete mission bundle — manifest, environment, checks and rubric — ready for the day ingestion ships.
- [Connect any agent](https://docs.xorcise.ai/guides/connect-your-agent.md): Point any AI agent at a XORCISE run: what registration records, what the connect prompt tells your agent, and how a run is correlated.
- [Send traces and logs](https://docs.xorcise.ai/guides/telemetry.md): Wire an agent's OpenTelemetry output into a XORCISE run, with the exact configuration for Claude Code, Codex, OpenHands, and everything else.
- [Track an agent over time](https://docs.xorcise.ai/guides/track-progress.md): Rank agents by their recorded results, read one agent's history, and compare scores across runs without drawing the wrong conclusion.
- [Introduction](https://docs.xorcise.ai/introduction.md): XORCISE evaluates cyber AI agents against realistic mission environments — locally, on your own machine.
- [Configuration](https://docs.xorcise.ai/operate/configuration.md): Where XORCISE keeps its settings, the handful worth changing, and how to upgrade or remove an install without orphaning anything.
- [Security and isolation](https://docs.xorcise.ai/operate/security.md): What XORCISE contains, what it does not, what leaves your machine, and how to run it safely on a shared host.
- [Troubleshooting](https://docs.xorcise.ai/operate/troubleshooting.md): Symptom-first fixes for the failures XORCISE actually produces at start, during a run, while grading, and when pulling missions.
- [REST API](https://docs.xorcise.ai/reference/api.md): Every XORCISE HTTP endpoint, with paths, parameters, bodies, response shapes, status codes, and the per-run token your agent authenticates with.
- [Checks and ops](https://docs.xorcise.ai/reference/checks.md): Every check source and op you can declare in a mission, the arguments each takes, and the exact conditions that make it pass or fail.
- [CLI reference](https://docs.xorcise.ai/reference/cli.md): Every xorcise command, option, exit code, and environment variable, with the web UI equivalent for each task.
- [Configuration keys](https://docs.xorcise.ai/reference/configuration.md): Every key XORCISE reads from config.toml or the environment, with its type, default, environment variable and precedence order.
- [Glossary](https://docs.xorcise.ai/reference/glossary.md): Every XORCISE term in one line, with a link to the page that defines it properly.
- [Mission manifest](https://docs.xorcise.ai/reference/mission-manifest.md): Every field of mission.json — the schema 2.0 manifest at the root of a mission bundle, for both lab and static missions.
- [Roadmap](https://docs.xorcise.ai/roadmap.md): What XORCISE does today, what carries an experimental label, and what is not built yet — with no dates attached.
- [Your first run](https://docs.xorcise.ai/start/first-run.md): Register the agent scout, evaluate it against chrono-canary, and finish holding a real score — the whole XORCISE loop, once.
- [Quickstart](https://docs.xorcise.ai/start/quickstart.md): Install XORCISE, start it, and open the web UI on your own machine in about five minutes.