Resources

Core references for setup, operator routes, CLI, agent skills, release history, and support.

by @saifyxpro

Primary references

ResourceWhat it covers
API Referencecurrent backend routes, methods, and queue behavior
CLIpublished headlessx command, login flow, and terminal usage
Agent Skillsinstall the repository skill into supported AI coding agents
MCP Setupremote MCP endpoint, auth, and client config
Changelogcurrent release history and migration context
Setup Guidelocal, mixed, and Docker runtime modes
API Endpointscurrent backend HTTP and MCP surface
GitHub Repositorysource, issues, and discussions

API surface at a glance

AreaRoute familyNotes
Health/api/healthpublic
Operator status/api/operators/statuscurrent operator catalog and availability
Website/api/operators/website/*HTML, JS HTML, content, screenshot, map, crawl, SSE
Google AI Search/api/operators/google/ai-search/*search, stream, status
Tavily/api/operators/tavily/*search, research, status
Exa/api/operators/exa/*search, stream, status
YouTube/api/operators/youtube/*extract, formats, subtitles, save, download
Jobs/api/jobs/*queue, stream reconnect, metrics, cancel
MCP/mcpremote MCP over HTTP

Use the dedicated API docs section for route-by-route detail:

Auth model

  • GET /api/health is public
  • every other /api/* route requires x-api-key
  • the dashboard uses DASHBOARD_INTERNAL_API_KEY on server-side proxy requests
  • MCP clients use only normal dashboard-created API keys

Streaming and jobs

Common SSE event names:

  • start
  • progress
  • result
  • error
  • done

Google AI Search currently ends with end instead of done.

Queue-backed features like crawl also require:

  • Redis
  • the worker process

Current external services

IntegrationEnvironment variable
TavilyTAVILY_API_KEY
ExaEXA_API_KEY
HTML-to-MarkdownHTML_TO_MARKDOWN_SERVICE_URL
YouTube engineYT_ENGINE_URL

When to use what

  • use the dashboard when you need fast visual validation
  • use HTTP routes for application code and workflow automation
  • use the CLI when you want markdown-first operator access from the terminal
  • use Agent Skills when a supported AI coding agent should drive the published CLI
  • use MCP for agent clients that benefit from remote tool discovery
  • use the changelog when you need release-specific context before upgrading

Related Docs

API Reference

Read the current backend route families, auth model, and dependencies.

Changelog

Review release history and migration context for older versions.

Next Steps

CLI

Install the published `headlessx` command and work against the operator API.

GitHub Repository

Explore the source code, report issues, and contribute to the project.