# HUD

## Docs

- [Converting Harbor Benchmarks](https://docs.hud.ai/advanced/harbor-convert.md): Import Harbor tasks into HUD with a single command
- [Advanced Patterns](https://docs.hud.ai/advanced/patterns.md): Sandboxing, mocking, scenario internals, and troubleshooting
- [Environments as Data](https://docs.hud.ai/building/environments-as-data.md): Design environments that produce useful training signal for reinforcement learning.
- [Deploy & Go Remote](https://docs.hud.ai/building/running-at-scale.md): Deploy your environment, sync tasks, and run evaluations at scale on HUD infrastructure.
- [Scaffolding](https://docs.hud.ai/building/scaffolding.md): Create an environment, define tools and scenarios.
- [Tasks & Evaluation](https://docs.hud.ai/building/tasks-and-evaluation.md): Define tasks, test locally, iterate, sync to the platform.
- [Changelog](https://docs.hud.ai/changelog.md): Product updates and release notes for HUD SDK and Platform.
- [Contributing](https://docs.hud.ai/contributing.md): Help us improve HUD
- [Build Your Own Codex](https://docs.hud.ai/cookbooks/codex-coding.md): Recreate OpenAI's Codex CLI from scratch using HUD
- [Build Your Own OpenCode](https://docs.hud.ai/cookbooks/opencode-agent.md): Recreate the OpenCode AI coding agent from scratch using HUD
- [Ops Diagnostics Agent](https://docs.hud.ai/cookbooks/ops-diagnostics.md): How we built a hierarchical agent to diagnose production issues across our infrastructure
- [Chat with Environments](https://docs.hud.ai/guides/chat.md): Build multi-turn conversational agents on top of HUD environments
- [Integrations](https://docs.hud.ai/guides/integrations.md): Use any agent framework with HUD environments
- [Introduction](https://docs.hud.ai/index.md): Build, evaluate, and train AI agents.
- [Add to Cursor/Claude](https://docs.hud.ai/llm-quickstart.md): Add context about hud to any coding agent
- [Automations](https://docs.hud.ai/platform/agents/automations.md): Run scenarios repeatably with pre-filled arguments — configure once, execute on demand
- [Chat Agents](https://docs.hud.ai/platform/agents/chats.md): Multi-turn conversational agents served over A2A — pair a scenario's tools with a model
- [QA Agents](https://docs.hud.ai/platform/agents/qa.md): Automated trace and task analysis — detect false negatives, false positives, reward hacking, and classify failures
- [Environments](https://docs.hud.ai/platform/environments.md): Deploy, manage, and monitor agent environments on the HUD platform.
- [File Tracking & Code Viewer](https://docs.hud.ai/platform/file-tracking.md): View file changes made during a trace in a full code viewer with syntax highlighting and diffs.
- [Overview](https://docs.hud.ai/platform/index.md): Navigate the HUD platform to manage models, environments, and tasksets.
- [Trace Analysis Environment](https://docs.hud.ai/platform/internal/trace-analysis.md): How we built an environment to let agents analyze their own traces—and how you can apply the same pattern
- [MCP Integration](https://docs.hud.ai/platform/mcp.md): Let your AI agent query the platform. Analyze traces, debug environments, explore tasks.
- [Models](https://docs.hud.ai/platform/models.md): Browse, fork, and train AI models on the HUD platform.
- [Publishing Leaderboards](https://docs.hud.ai/platform/publishing-leaderboards.md): Create a taskset, run evaluations, validate results, and publish to the community.
- [REST API](https://docs.hud.ai/platform/rest-api.md): Programmatically run scenarios and retrieve trace results via the HUD REST API
- [Slack Integration](https://docs.hud.ai/platform/slack.md): Run HUD scenarios directly from Slack by @mentioning the HUD bot.
- [Subagent API](https://docs.hud.ai/platform/subagent.md): Create a scenario, optionally train a model, and get a simple API endpoint that runs your agent.
- [Tasksets](https://docs.hud.ai/platform/tasksets.md): Organize and manage collections of tasks for agent evaluations.
- [Models](https://docs.hud.ai/quick-links/models.md): One endpoint for every model. Native tool integration.
- [Evaluations & Training](https://docs.hud.ai/quick-links/training.md): Run evaluations at scale. Train models on your tasks.
- [Agents](https://docs.hud.ai/reference/agents.md): SDK reference for HUD agent classes
- [hud analyze](https://docs.hud.ai/reference/cli/analyze.md): Inspect MCP environments quickly from metadata or live containers
- [hud build](https://docs.hud.ai/reference/cli/build.md): Build production images and generate lock files for reproducible environments
- [hud debug](https://docs.hud.ai/reference/cli/debug.md): Test MCP environments through 5 validation phases
- [hud deploy](https://docs.hud.ai/reference/cli/deploy.md): Deploy HUD environments directly to the platform
- [hud dev](https://docs.hud.ai/reference/cli/dev.md): Development server for MCP environments (hot-reload is opt-in via --watch)
- [hud eval](https://docs.hud.ai/reference/cli/eval.md): Run agents on tasks or datasets
- [hud init](https://docs.hud.ai/reference/cli/init.md): Create a new HUD environment from a preset
- [hud link](https://docs.hud.ai/reference/cli/link.md): Link a local directory to an existing HUD environment
- [Other CLI Commands](https://docs.hud.ai/reference/cli/misc.md): Miscellaneous HUD CLI utilities
- [CLI Overview](https://docs.hud.ai/reference/cli/overview.md): Complete reference for HUD command-line tools
- [hud push](https://docs.hud.ai/reference/cli/push.md): Share HUD environments to Docker and HUD registries
- [hud rl](https://docs.hud.ai/reference/cli/rl.md): Submit and monitor RL training jobs
- [hud sync](https://docs.hud.ai/reference/cli/sync.md): Push local task definitions to the HUD platform for team evaluation and CI
- [Environment](https://docs.hud.ai/reference/environments.md): SDK reference for the Environment class - tools, connectors, and integrations
- [Evals](https://docs.hud.ai/reference/evals.md): SDK reference for hud.eval() - the unified evaluation context manager
- [Native Graders](https://docs.hud.ai/reference/native-graders.md): SDK reference for reusable first-party evaluation helpers in hud.native
- [Tools](https://docs.hud.ai/reference/tools.md): SDK reference for building and extending HUD tools
- [Types](https://docs.hud.ai/reference/types.md): SDK type reference for HUD
- [AgentTool](https://docs.hud.ai/tools/agents.md): AgentTool — run sub-agents as callable tools for hierarchical task delegation
- [Coding Tools](https://docs.hud.ai/tools/coding.md): Shell execution and file editing
- [Computer Tools](https://docs.hud.ai/tools/computer.md): Mouse, keyboard, and screenshot control
- [Filesystem Tools](https://docs.hud.ai/tools/filesystem.md): File reading, searching, and directory listing
- [Grounding Tools](https://docs.hud.ai/tools/grounding.md): Element descriptions to coordinates
- [Memory Tools](https://docs.hud.ai/tools/memory.md): Persistent storage across conversations
- [Web Tools](https://docs.hud.ai/tools/web.md): Browser automation and web search

## OpenAPI Specs

- [openapi](https://docs.hud.ai/api-reference/openapi.json)