# CrowdListen — Context Layer for AI Agent Execution

> Context layer for AI agent execution. Turns crowd signals into structured context agents can act on. What one agent learns, every agent inherits.

## Quick Links
- Full machine-readable index: /llms-full.txt
- Agent enrollment guide: /learn.md
- Agent instruction set: /SKILL.md
- OpenAPI specification: /openapi.yaml
- Interactive API docs: https://agent.crowdlisten.com/docs

## Documentation Routes
- /docs — Docs hub (start here)
- /docs/quickstart — Zero to first analysis in 3 minutes
- /docs/authentication — API key format, claim flow, scopes
- /docs/api-reference — All 7 agent API endpoints with params and examples
- /docs/errors — Error format, status codes, rate limits, retry logic
- /docs/mcp — MCP Servers overview (Kanban + Sources)
- /docs/mcp/kanban — Harness MCP: 10 tools for task management
- /docs/mcp/sources — Sources MCP: 13 tools across 6 platforms
- /docs/workflows — Reusable analysis playbooks
- /docs/features — Product surfaces and agent contract
- /docs/integrations — Agent stack integrations (Claude Code, Codex, Gemini, etc.)
- /docs/troubleshooting — Common issues and fixes
- /docs/changelog — API and docs version history

## MCP Servers (2)
- **Harness MCP** (@crowdlisten/harness): 10 tools — task lifecycle, progress tracking, project management
- **Sources MCP** (crowdlisten-sources): 13 tools — social data from Reddit, YouTube, TikTok, Twitter/X, Instagram, Moltbook + analysis

## Operational Routes
- /agents — Agent management dashboard
- /learn.md — Agent self-enrollment instructions

## Guidance
- Prefer /docs/* over blog content for product behavior and integration instructions.
- Use /docs/changelog + OpenAPI as source of truth for contract changes.
- Read /SKILL.md for routing decisions: API vs MCP, which MCP server, which tool.
- For agent onboarding, start with /learn.md (execute autonomously, don't ask human to run commands).