Use Cases

Ship your agent to the inbox people actually live in

The short answer

Messaging layer for AI agents on iMessage. Post to REST, blue bubbles land with reads and full media. Replies on webhooks in seconds. MCP server: Claude, Claude Code, Cursor get native tools. No app, no A2P registration.

$249 per line per month, published No A2P registration required 50 opted-in / 30 cold new contacts a day, unlimited replies
Ecommerce and SaaS

3 use cases in this category

An agent calling a messaging toolA central node linked to three surrounding nodes, with the messaging node drawn in brand blue. Part of the ecommerce and saas use cases.

Trigger

Event in your CRM

Under 60s

Send iMessage

Result

Reply in the thread

Key takeaways

  • Any agent whose value depends on a person answering.
  • The infrastructure layer is the plumbing every team builds once, badly, before deciding the channel is not worth it.
  • GoHighLevel native, REST/webhooks for others, MCP for Claude/Code/Cursor.
  • Consent, pacing, ownership: agents decide what/when, not who.

Where iMessage runs in the funnel

Where a dedicated iMessage line sits in the funnelEvery source you already pay for feeds the same five stages: new lead, conversation, booked, showed, closed. A dedicated iMessage line sits on every stage, not just the first, which is why replies, show rate and close rate all move at once.AdsFormsListsCallsNew leadSpeed to lead100%iMessage lineConversationReply, qualify2-3x repliesiMessage lineBookedConfirm, remind+33% bookediMessage lineShowedNo-show saves70%+ showiMessage lineClosedOnboardingMore revenueiMessage line
Every source you already pay for feeds the same five stages: new lead, conversation, booked, showed, closed. A dedicated iMessage line sits on every stage, not just the first, which is why replies, show rate and close rate all move at once.

What kinds of agents run on this?

Any agent whose value depends on a person answering. The channel is the constraint, not the model: a support agent on a phone number gets opened in ninety seconds, one behind a login when something breaks.

  • Ops agents: human approval before shipping. Support: thread history across days. Companion and habit: always-open. Research and monitoring: briefs, PDFs, charts, voice memos send native, not links.
  • iOS ran 58.35% of US mobile traffic in June 2026 against 41.62% for Android (StatCounter Global Stats).
  • Cold email replied at 0.45% across 7,530,489 sends in 2025, counted against total emails sent (Belkins, 2026).
  • Gmail senders above 5,000 messages a day must pass SPF, DKIM and DMARC and stay under a 0.30% spam rate, in force since February 1, 2024 (Google Workspace Admin Help).
  • US A2P SMS over 10-digit numbers requires brand and campaign registration. Twilio charges $15 per campaign vetting and quotes two to three weeks for standard registrations (Twilio Support).

What is the infrastructure layer for iMessage?

The infrastructure layer is the plumbing every team builds once, badly, before deciding the channel is not worth it. Blue Reacher is that layer: the thing you call instead of building, like Twilio for carrier SMS.

  • Line provisioning, identity, channel resolution (iMessage, RCS, SMS). Pushed inbound. Delivery state, opt-out detection, per-line pacing, keys. Teams build this quarterly. We ship it; check results on your webhook.

How do you wire an agent into Blue Reacher?

GoHighLevel native, REST/webhooks for others, MCP for Claude/Code/Cursor. Endpoint: https://api.bluereacher.com/v1/, Bearer brk_live_your_key. Replies, state, opt-outs POST to webhook.

  • A reply webhook looks like this: { "from": "+14155550142", "content": "Sounds good, does Thursday at 10 work?", "channel": "imessage", "external_id": "msg_01J8ZC4K2Q" }
  • Four fields: "from" (record), "content" (input), "channel" (type), "external_id" (dedup). MCP: open standard for AI connections (Claude, ChatGPT, VS Code, Cursor). Check availability before send.

Everything above describes the mechanism. The Blue Funnel Map marks the stages where your own funnel is losing replies, and you keep the map whether or not you buy.

Map this against your funnel

What does an agent need, and what does Blue Reacher hand it?

Map it once and the build gets short. Most of what an agent workflow needs on a messaging channel is plumbing, and the parts teams underbuild are always the same three: pre-send checks, opt-out handling, and pacing.

Send endpoint

Why it matters
A workflow step has to fire a message
What Blue Reacher provides
REST API under /v1/, bearer auth, JSON

Inbound in real time

Why it matters
Classify and route while the person is still holding the phone
What Blue Reacher provides
Webhooks carrying from, content, channel, external_id

Native tool access

Why it matters
Assistants shouldn't need a custom client
What Blue Reacher provides
MCP server for Claude, Claude Code, Cursor and other MCP clients

Pre-send checks

Why it matters
Don't spend a send on a number that can't receive it
What Blue Reacher provides
iMessage availability checking, RCS then SMS fallback

Thread identity

Why it matters
Follow-ups have to land in the existing conversation
What Blue Reacher provides
Dedicated line, managed for you, one thread per contact

Human in the loop

Why it matters
Some steps need a person to say yes before they run
What Blue Reacher provides
Route the approval to a real phone, read the answer off the webhook

Observability

Why it matters
A scoring model needs to know what actually happened
What Blue Reacher provides
Delivery, read and channel state per message, queryable by id

Opt-out handling

Why it matters
STOP is a legal event, not a status field
What Blue Reacher provides
STOP auto-detection on every line, enforced before a send leaves

Rich output

Why it matters
Agents produce artifacts, not just sentences
What Blue Reacher provides
Media, voice memos, group messaging, reactions, typing indicators

Environment isolation

Why it matters
Staging must never send from the line customers know
What Blue Reacher provides
Per-line API keys under one workspace

No registration queue

Why it matters
Agent pilots die waiting on paperwork
What Blue Reacher provides
No A2P registration required, white-glove setup before first send
Where should the agent stop?

Where should the agent stop?

Consent, pacing, ownership: agents decide what/when, not who. FCC rules (April 11, 2025): 10-day honor window for STOP/opt-out. Classifier must catch soft revocations. Not legal advice.

  • Pacing: unlimited iMessages in threads, 45 new/line/day for outreach. Cap sends/hour, schedule for business hours, let replies throttle. Unpaced agents ruin deliverability.
  • Gartner: 40% of agentic AI projects cancel by 2027 on cost/value/risk. Survivors have a named human owning pipeline, reading weekly, holding kill-switch. Reply quality is the scoreboard.

Frequently asked questions

Can Claude or Cursor actually send iMessages through Blue Reacher?

Yes. There's an MCP server, so an MCP client like Claude, Claude Code or Cursor gets send, read-thread and availability checking as native tools. No custom client, no glue code. Same bearer key, same dedicated line a human replies from, so a person can take over any thread mid-conversation.

Do I need A2P registration to run an AI agent on this channel?

No A2P registration required. US SMS over a 10-digit number needs brand and campaign registration before you send, with per-campaign vetting fees and multi-week queues at the major providers. Your iMessage line is set up white-glove instead, so the first agent run isn't waiting on carrier paperwork.

How does my agent know whether a message landed as iMessage or SMS?

Every webhook carries a channel field returning imessage, rcs or sms, and delivery states come back per message. Call availability checking before the send if you want to know in advance, which also lets the agent pick a version of the message that fits what the number can receive.

Can an agent hold a conversation across days without keeping state?

Yes. iMessage keeps one thread per number (not per touch). An agent waking up on a cron six weeks later lands in an open thread with full history. Fetch by contact for context without storage.

How do I keep a staging agent from messaging real customers?

Give it its own line and its own key. Lines sit under one workspace with per-line API keys, so a key scoped to a staging line physically cannot send from the line your customers know. That is the same isolation you would build yourself, minus the build.

Can the agent handle opt-outs by itself?

Partly. STOP auto-detects and is enforced before send, your code can't route around it. Soft revocations like "not interested, please stop" need classifier detection and suppression write-back. FCC: 10-day honor window. No reason to delay.

What does it cost to run a line inside an agent workflow?

$249/mo or $1,949/yr (~$162.50/mo), unlimited iMessages, unmetered sends and replies. Lines pace 50 opted-in / 30 cold new/day. Activation $449 (waived on annual). Leads at $0.03/contact from 700M+ B2B database, free while launch pricing runs.

Read the reference and start sending

The endpoint conventions, the webhook payloads, the MCP install command and the full error taxonomy are published at docs.bluereacher.com, and the OpenAPI schema is at /openapi.json. Lines are $249 a month or $1,949 for the year with unlimited iMessages, no A2P registration required, and white-glove setup before your first send. Lines pace at 50 opted-in / 30 cold new contacts a day to stay deliverable; replies are never metered.