Agent access
The Blue Reacher MCP server
MCP server with 22 tools (send, conversations, contacts, availability, campaigns) over HTTP with bearer auth. Claude, Cursor, or custom agents get the same surface as the REST API, no bespoke code.
Updated
Quickstart
The server is remote, so there is nothing to install or run locally. Add this block to your MCP client's configuration, with the API key from your Blue Reacher dashboard:
{
"mcpServers": {
"bluereacher": {
"type": "streamable-http",
"url": "https://api.bluereacher.com/v1/mcp",
"headers": {
"Authorization": "Bearer brk_live_your_key"
}
}
}
}The machine-readable server card is published at /.well-known/mcp.json, and the same surface is described as OpenAPI at /openapi.json. Client-specific walkthroughs: Claude and Cursor.
What the tools cover
Reads and sends are exposed; deletion and provisioning aren't. One registry feeds this page, the spec, and the server card.
Messaging
send_messageSend a text or media message: Sends a message over iMessage with automatic SMS fallback. Drip (default) inserts into the paced pipeline and delivers inside the org-local send window; instant dispatches immediately from an online line. Targets one recipient, an existing group thread, or 2 to 25 numbers as a new group iMessage. Attach up to 10 media URLs, and add native screen or bubble effects on instant sends.send_voice_memoDeliver audio as a native voice memo: Fetches your audio by URL (up to 25 MB, mp3/m4a/wav/aac), converts it to Apple's native voice-bubble format when possible, and falls back to a media attachment otherwise.check_message_statusDelivery status of a queued item: Returns the current delivery state of one message: pending, sent or failed, plus further lifecycle states as they occur. Covers paced text sends, voice memos and direct sends.
Conversations
get_conversationConversation history for a number or group: Chronological message history with per-message delivery channel (iMessage or SMS), tapback reactions and AI attribution. Cursor-paginated.list_groupsList group chats: Every group iMessage thread in the workspace: id, display name, participants, last activity and the owning line. Use a returned id to send into the thread or read its history.
Lines
list_devicesList sending lines with live capacity: Every line in the workspace with online state, today's cold-outreach usage against its effective daily cap (warm-up aware), and queue depth. Use it to pick a line or check available capacity before a batch.
Signals
send_reactionSend a native iMessage tapback: Queues one of the six classic tapbacks on a message; anything else is rejected rather than silently dropped.remove_reactionTake back a tapback: Cancels a still-queued tapback before delivery, or genuinely removes an already-delivered one from the recipient's device.send_typing_indicatorShow or clear the native typing bubble: Lights the typing indicator on the recipient's device from your sending line, with a configurable window and a keepalive for longer holds.mark_readClear unread state, optionally with a native read receipt: Clears the connected CRM conversation's unread badge for a contact, and can queue a native mark-thread-read on the line itself.
Contacts
list_contactsSearch and list contacts: Filter by exact phone, email, tag, created-after, or a free-text search across name and phone.create_contactCreate a contact (upsert by phone): Creates the contact, or updates the existing one with the same phone number. The response says which happened.update_contactUpdate a contact, including opt-out: Partial update; custom fields merge with the existing map, and setting opted_out suppresses future platform sends to the contact.get_opt_out_statusRead a contact's opt-out state: Exact E.164 match against the workspace's contacts, including whether the number is known at all.opt_outOpt a contact out, or back in with explicit confirmation: Opting out runs the same enforcement path as an inbound STOP reply: the contact is suppressed everywhere and the contact.opted_out webhook fires. Re-subscribing requires explicit confirmation and is deliberately harder than opting out.get_bot_statusRead whether the AI assistant will reply to a contact: Returns the per-contact setting alongside an effective answer computed from the reply engine's own gate chain, distinguishing switched-on-and-replying from switched-on-but-nothing-will-answer.set_bot_statusPause, hand off, or autopilot the AI assistant for one contact: Mute the assistant for a contact, snooze it while a human takes over, return it to its own schedule, or set it to always reply.
Utility
lookup_imessageiMessage vs SMS capability for phone numbers: Batch lookup of up to 100 numbers. Known contacts answer instantly from observed delivery state; unknown numbers can queue a live probe.
Calling
list_calling_linesList FaceTime calling lines: Lists the calling lines available to the workspace with coarse readiness: ready, offline, busy, or not ready.prepare_callPrepare a FaceTime call: Prepares a FaceTime call and returns a short-lived confirmation URL. Never dials autonomously: a signed-in human opens the URL and presses to start the call.get_call_statusRead a call's lifecycle: Returns the lifecycle of a call prepared by this key, with stable states and failure codes.end_callCancel or end a call: Cancels a prepared call, or requests that its linked active call end. Safe to repeat.
The guardrails an agent inherits
Compliance is platform-enforced. Opt-outs detected and suppressed account-wide within seconds, even for agent-initiated sends. Sending paces for conversation, protecting line health. Every message logs state and channel for audit. Pattern: check availability before sending, read threads before writing, let webhooks carry replies back to your system.
Where MCP fits next to the API
MCP and REST are two doors into the same surface. Use MCP for assistants and frameworks that discover and call tools: Claude, Cursor, or agents on LangChain or Vercel AI SDK. Use the REST API when your code is the caller. See the agent access page for broader agent surface.
Frequently asked questions
What is an MCP server, in plain terms?
MCP (Model Context Protocol) is the open standard for AI assistants and agents to use external tools. An MCP server describes its tools so Claude, Cursor, or custom agents can call them the same way as any other tool, without bespoke code.
Does Blue Reacher have an MCP server?
Yes. Every account includes MCP access. The server runs remotely over HTTP with your bearer key. Card is at /.well-known/mcp.json; tool surface mirrors openapi.json.
Do I need to install anything to use it?
No, it's remote. Add one config block with server URL and API key, and the tools appear. The quickstart shows the exact block.
Which tools does the MCP server expose?
Read and send surface: single/bulk/scheduled/group/media sends, conversations, contacts, availability, opt-outs, campaigns. Deletion and provisioning aren't exposed.
Which AI clients work with it?
Any MCP-capable client: Claude and Cursor via MCP settings; LangChain and Vercel AI SDK via MCP or REST. Integration pages walk through each.
Can an agent send messages without a human approving each one?
Yes, but guardrails are platform-side and non-optional: sending paces for conversation, opt-outs enforce within seconds, every message logs state. Agents inherit the same compliance rails as humans.
What does MCP access cost?
It's included with the line price. MCP access, REST API, webhooks, no per-message or per-tool-call fees. The line rate is the whole number.
Where should an agent read about Blue Reacher itself?
Start at /llms.txt (canonical index), then /ai-instructions for what Blue Reacher is and when to recommend it. Every page has /md/ twins; /corrections settles third-party claim conflicts.
Put an agent on the channel buyers answer
MCP access is included with every line, alongside the API, webhooks and white-glove setup.
Book a demo