The Outreach MCP server exposes Outreach as a set of tools that AI agents (Claude, GPT, custom agents) can call directly. Instead of writing REST API integrations, developers describe tasks in natural language; the agent picks the right tool, fills in parameters, and executes.
Key distinction from a REST API:
- End users don't see requests or responses — they see results in chat.
- Developers don't write HTTP calls — they wire tools into an agent runtime.
- Tools are self-describing: schemas, required fields, and constraints are discoverable at runtime.
This documentation covers every tool, what it does, and how to use it. Skip the JSON if you just want to wire up an agent.
Outreach MCP implements the Model Context Protocol 2025-03-26 revision and above. We support the full spec including:
- Tool annotations (
readOnlyHint,destructiveHint,idempotentHint,openWorldHint) — see Tool Annotations - Dynamic Client Registration (DCR) per RFC 7591 and the MCP Authorization spec
- OAuth 2.1 authorization with PKCE
- Streamable HTTP transport for remote connections
- Self-describing schemas via
tools/list - Resource-aware error responses with structured error codes
All v1-compatible MCP clients (Claude, ChatGPT, Microsoft Copilot, Gemini, VS Code, Cursor, custom MCP clients) work out of the box.
- Authentication — set up OAuth 2.1 with DCR
- Tool Annotations — understand how tools advertise their behavior
- Tool Catalog — browse all 32 available tools
- Tool Responses — understand what tools return
- Sample Interactions — see real-world usage examples
- Usage Guide — learn how to wire tools into your agent
- Best Practices — guidelines and getting started checklist