Ship your agent to the inbox people actually live in
The short answer
REST API, MCP server, and webhooks for AI agents on iMessage. Blue bubbles land with reads and media. Replies fire on your webhook. Claude, Claude Code and Cursor get native tools.
3 use cases in this category
Trigger
Event in your CRM
Under 9 seconds
Send iMessage
Result
Reply in the thread
Key takeaways
- Any agent whose value depends on a person answering in minutes, not hours.
- Provisioning, identity, channel detection, webhooks, delivery state, per-line pacing, keys.
- GoHighLevel is native. Others use REST API (https://api.bluereacher.com/v1/) with Bearer auth, or MCP tools if your agent is Claude or Cursor.
- Three parts teams always underbuild: pre-send checks, opt-out detection, per-line pacing.
- Agents decide what and when; compliance decides who.
Where iMessage runs in the funnel
What kinds of agents work on iMessage
Any agent whose value depends on a person answering in minutes, not hours. Support agents on a phone get opened in 90 seconds; the same agent behind a login in a dashboard gets checked once a day.
- Ops agents: one-step approvals before shipping.
- Support: thread history across days with one contact.
- Companion: always-on in the app people live in.
- Research: briefs, PDFs, charts, voice memos send native, not links.
What an agent platform needs to provide
Provisioning, identity, channel detection, webhooks, delivery state, per-line pacing, keys. Most teams build twice, badly. Blue Reacher is what you call instead.
Three ways to integrate: REST, MCP, or native
GoHighLevel is native. Others use REST API (https://api.bluereacher.com/v1/) with Bearer auth, or MCP tools if your agent is Claude or Cursor. Replies, delivery, opt-outs POST to your webhook.
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 funnelThe plumbing that saves months of rebuilds
Three parts teams always underbuild: pre-send checks, opt-out detection, per-line pacing. Map once, build short.
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
- Signed webhooks carrying phone_number, content, service_type and event_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
Guardrails: where agents draw the line
Guardrails: where agents draw the line
Agents decide what and when; compliance decides who. Unlimited iMessages in threads; outreach paced 50 opted-in/30 cold per day. FCC: 10-day opt-out. Human reviews weekly with kill-switch.
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.
Ship your agent to the inbox people actually read
Read the OpenAPI spec at docs.bluereacher.com/openapi.json. Lines are $249 a month or $1,949 per year with unlimited iMessages, no A2P registration, and white-glove setup. Book a walkthrough.