Documentation
MCP Server
Connect AI agents to OutreachAgent via the Model Context Protocol (MCP). The @outreachagent/mcp package exposes all 75 API operations as MCP tools, letting agents in Cursor, Claude Desktop, or any MCP-compatible client manage outreach autonomously.
OutreachAgentClient method via stdio transport.Architecture
AI Agent (Cursor / Claude Desktop / custom MCP client)
│
└── stdio (JSON-RPC) ──► @outreachagent/mcp
│
└── @outreachagent/sdk-ts
│
└── HTTPS ──► api.outreachagent.devQuick Start
Cursor
Add the following to your project's .cursor/mcp.json:
{
"mcpServers": {
"outreachagent": {
"command": "npx",
"args": ["@outreachagent/mcp"],
"env": {
"OUTREACHAGENT_API_KEY": "rm_live_..."
}
}
}
}Claude Desktop
Add the following to your claude_desktop_config.json:
{
"mcpServers": {
"outreachagent": {
"command": "npx",
"args": ["@outreachagent/mcp"],
"env": {
"OUTREACHAGENT_API_KEY": "rm_live_..."
}
}
}
}Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
OUTREACHAGENT_API_KEY | Yes | — | Your OutreachAgent API key |
OUTREACHAGENT_BASE_URL | No | https://api.outreachagent.dev | API base URL (for self-hosted or development) |
Available Tools (75)
Every method on the TypeScript SDK is exposed as an MCP tool. Tools are named with kebab-case and grouped by resource:
| Group | Tools | Count |
|---|---|---|
| Organizations | list-organizations | 1 |
| API Keys | list-api-keys, create-api-key, revoke-api-key | 3 |
| Pods | list-pods, get-pod, create-pod, delete-pod | 4 |
| Inboxes | list-inboxes, get-inbox, create-inbox, update-inbox, delete-inbox | 5 |
| Messages | list-messages, get-message, send-message, search-messages | 4 |
| Contacts | list-contacts, get-contact, create-contact, update-contact, delete-contact | 5 |
| Templates | list-templates, get-template, create-template, update-template, delete-template, preview-template | 6 |
| Workflows | list-workflows, get-workflow, create-workflow, update-workflow, delete-workflow, publish-workflow, pause-workflow, resume-workflow, simulate-workflow, test-send-workflow, preview-workflow | 11 |
| Enrollments | list-enrollments, get-enrollment, create-enrollment, list-enrollment-logs | 4 |
| Policies | list-policies, create-policy, update-policy, delete-policy | 4 |
| Approvals | list-approvals, get-approval, approve-request, reject-request | 4 |
| Threads | list-threads, get-thread | 2 |
| Domains | list-domains, get-domain, create-domain, delete-domain | 4 |
| Webhooks | list-webhook-endpoints, get-webhook-endpoint, create-webhook-endpoint, update-webhook-endpoint, delete-webhook-endpoint, list-webhook-deliveries, replay-webhook-delivery, list-events | 8 |
| Segments | list-segments, get-segment, create-segment, update-segment, delete-segment | 5 |
| Metrics | get-metrics-summary | 1 |
| Extractions | extract-otp, extract-invoice-basic | 2 |
| Realtime | create-realtime-session | 1 |
| Billing | create-checkout-session | 1 |
Example: Agent-Driven Outreach
Once the MCP server is configured, an AI agent can autonomously run a cold outreach campaign using natural language. For example, an agent in Cursor could:
- Call
create-contactto add prospects with attributes like company, title, and role. - Call
create-templateto author personalized email templates with Liquid variables. - Call
create-workflowto build a multi-step sequence with delays, follow-ups, and exit-on-reply. - Call
publish-workflowto activate the workflow. - Call
create-enrollmentto enroll each contact into the workflow. - Call
get-metrics-summaryto monitor delivery rates and adjust strategy.
requires_approval response and can call approve-request or wait for human review.Error Handling
All tool errors are returned as structured MCP error responses with isError: true. The error message includes the HTTP status code, error code, and human-readable description from the OutreachAgent API.
Error 404 (not_found): Contact not found
Links
- TypeScript SDK Reference — Full method signatures and types
- OpenAPI Spec — Machine-readable API specification
- LLM Context — Complete API reference optimized for LLM context windows
- Model Context Protocol — MCP specification and ecosystem