> ## Documentation Index
> Fetch the complete documentation index at: https://docs.reply.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Reply MCP overview

> The official, live Reply.io MCP server — 70 tools for sequences, contacts, inbox, tasks, and the Jason AI SDR autopilot.

Reply MCP is an official, live, remote [Model Context Protocol](https://modelcontextprotocol.io)
server that exposes Reply.io's sales-engagement, inbox, AI SDR (Jason), task, account, and help
workflows as **70 tools** for any MCP-compatible client — Claude, Cursor, Make, n8n, and custom agents.

<Note>
  Everything in these pages was read off the running server via an authenticated
  `initialize` + `tools/list` handshake and safe validation probes. Tool names, signatures,
  enum values, annotations, response envelopes, and error codes come from the live server —
  not hand-maintained prose. If any other documentation disagrees, trust the server (see
  [Regenerate this reference](/mcp/tools#regenerate-this-reference)).
</Note>

## At a glance

|                  |                                                                                  |
| ---------------- | -------------------------------------------------------------------------------- |
| **Endpoint**     | `https://mcp.reply.io/`                                                          |
| **Transport**    | Streamable HTTP (responses are SSE `text/event-stream` frames carrying JSON-RPC) |
| **Wire format**  | JSON-RPC 2.0 over HTTP                                                           |
| **MCP protocol** | `2025-06-18`                                                                     |
| **Auth**         | `Authorization: Bearer <key>`                                                    |
| **Rate limit**   | Hourly window, \~3000/hour (`X-Rate-Limit-*` headers on every response)          |
| **Sessions**     | None required — `initialize` and `tools/list` work without an `Mcp-Session-Id`   |
| **Tools**        | 70 — 31 annotated `readOnlyHint`, 39 annotated `destructiveHint`                 |
| **Capabilities** | `tools`, `logging` (no prompts or resources advertised)                          |

<CardGroup cols={2}>
  <Card title="Connect your client" icon="plug" href="/mcp/connect">
    Add Reply MCP to Claude, Cursor, or any MCP client in a couple of minutes.
  </Card>

  <Card title="Tool reference" icon="wrench" href="/mcp/tools">
    The full catalog of all 70 tools, grouped by domain.
  </Card>

  <Card title="Tool contract & errors" icon="file-contract" href="/mcp/tool-contract">
    The rules every tool enforces: schemas, envelopes, error codes, retries.
  </Card>

  <Card title="Building agents" icon="robot" href="/mcp/agent-guide">
    Design principles, high-stakes gating, and a drop-in system prompt.
  </Card>

  <Card title="Guides: what to build" icon="map" href="/guides/launch-outreach">
    Task-based guides — launch outreach, process replies, build an AI SDR.
  </Card>

  <Card title="Choose your interface" icon="scale-balanced" href="/agents/choose-your-interface">
    Not sure MCP is the right surface? Compare MCP, Skills, CLI, and API.
  </Card>
</CardGroup>

## What's exposed

The 70 tools cover the day-to-day Reply.io surface an agent needs:

| Domain                    | Tools | What it does                                                                                             |
| ------------------------- | :---: | -------------------------------------------------------------------------------------------------------- |
| **Sequences**             |   16  | Discover, inspect, start/pause, enroll contacts, and configure mailboxes, schedules, and AI SDR settings |
| **Contacts**              |   8   | Look up, filter, create, update, reassign, blacklist, and read activity                                  |
| **Inbox & conversations** |   3   | List inbox threads, send replies, categorize                                                             |
| **Jason AI SDR**          |   31  | Approvals & autopilot, knowledge bases, reply handlers, reengagement cards, offers, and playbooks        |
| **Tasks**                 |   3   | List, create, and complete tasks                                                                         |
| **Workspace resolvers**   |   5   | Resolve mailboxes, LinkedIn accounts, schedules, lists, and teammates to their IDs                       |
| **Help & meta**           |   4   | Search the Help Center, read articles, map the app, report unsupported requests                          |

See the [tool reference](/mcp/tools) for every tool name, its safety annotation, and a one-line purpose.

## MCP vs the REST API

MCP and the [REST API](/api-reference/introduction) are two surfaces over **one product** — same
account, same credit system.

* **MCP** — the interactive surface. 70 curated, annotated tools for agent actions and
  natural-language workflows. Best for day-to-day operations.
* **REST** (`https://api.reply.io/v3`) — the exhaustive surface. Use it for bulk imports and
  updates, background jobs, deep report exports, webhooks, and template/schedule/mailbox
  management not exposed as tools.

<Card title="REST API reference" icon="code" href="/api-reference/introduction">
  Browse the full v3 REST API — the exhaustive surface for bulk and background operations.
</Card>

## FAQ

<AccordionGroup>
  <Accordion title="Is the MCP server real, and how current is this page?">
    Yes — it's an official, live server at `mcp.reply.io`. These pages were generated from an
    authenticated `initialize` + `tools/list` handshake against the running server. If the
    server ever returns a different tool set, the server is newer — trust it, and
    [regenerate the reference](/mcp/tools#regenerate-this-reference) in seconds.
  </Accordion>

  <Accordion title="How does authentication work?">
    Pass your Reply.io API key as an `Authorization: Bearer` header. A call with
    no credentials returns `401` with a `WWW-Authenticate` challenge; OAuth protected-resource
    metadata is also published for clients that prefer a discovery-based flow. See
    [Connect](/mcp/connect).
  </Accordion>

  <Accordion title="How do I know which tools are safe to call?">
    Every tool is annotated in `tools/list`: 31 are `readOnlyHint: true` and 39 are
    `destructiveHint: true`. Agents can — and should — gate on these annotations before adding
    their own confirmation UX for the high-stakes subset. See the
    [tool contract](/mcp/tool-contract).
  </Accordion>

  <Accordion title="What do responses look like?">
    Responses are SSE frames; read the last `data:` frame. A successful `tools/call` returns the
    payload as a JSON string in `result.content[0].text`. Tool failures are **not** HTTP errors —
    they arrive as HTTP 200 with `result.isError = true`. See
    [response envelopes](/mcp/tool-contract#response-envelopes).
  </Accordion>

  <Accordion title="When should an agent use MCP instead of the REST API?">
    Use MCP for interactive, natural-language workflows and everyday operations. Drop to the
    REST API for bulk imports/updates, background jobs, report exports, and anything absent from
    `tools/list`.
  </Accordion>

  <Accordion title="Is the Jason AI SDR autopilot really callable here?">
    Yes. 31 of the 70 tools drive Jason — the approval queue, knowledge bases, reply handlers,
    reengagement cards, offers, playbooks, and the Review/Autonomous reply mode.
  </Accordion>
</AccordionGroup>