Agentic API

Build AI agents on iMessage.

Q: Can I build a fully autonomous agent with Tuco?

Yes. Set up a webhook to receive replies, pipe them through your AI model (Claude, GPT, custom), and send the response back through Tuco's API. The loop runs 24/7 without human intervention. Most customers add a confidence threshold to escalate to a human when the AI isn't sure.

Q: What about Apple's Terms of Service?

Tuco runs on real Apple hardware — dedicated iPhones and Macs, not emulators or VMs. Messages send through native iMessage infrastructure the same way you'd send from your personal device. We stay fully compliant.

Q: How fast are webhooks for agent reply loops?

Webhooks fire within 2-3 seconds of the event. For an AI agent loop — customer replies, webhook fires, AI processes, Tuco sends reply — the total round-trip is typically under 5 seconds. Feels like texting a real person.

Q: How do I prevent my agent from sending walls of text?

Tuco's system works with messages, not emails. Our recommended approach: include a system prompt that enforces message-appropriate length (under 160 characters for texts). Some customers also use our proprietary text refinement that compresses AI output into natural, concise messages.

Q: Can my agent send images, PDFs, and voice notes?

Absolutely. Tuco supports images, videos (up to 25MB), PDFs, contact cards, and audio files. Your agent can send a product photo, a contract PDF, or a voice memo — all natively through iMessage.

Q: What happens when my AI agent can't handle a question?

Build a confidence threshold into your agent logic. When the AI's confidence drops below your threshold, route the conversation to a human via Slack, your CRM, or a custom dashboard. Tuco handles the messaging layer — you handle the routing.

Your AI agent needs to talk to humans. Email is too slow. SMS gets filtered. iMessage is native, trusted, and read within 3 minutes. Tuco gives your agent the communication layer it's been missing.

Build Your Agent View API Docs

Works with Claude, GPT, Gemini, LLaMA, or any custom model

The Problem

AI agents are smart. But they can't reach anyone.

Your agent can reason, retrieve, and respond. But the last mile — actually reaching the human — is broken.

📧

Email: too slow.

Average response time is 4+ hours. Your agent's perfectly crafted reply sits unread in a promotions tab. By the time they see it, the moment's gone.

📴

SMS: filtered.

Carriers flag automated SMS. 10DLC registration takes weeks. Throughput caps kill your agent's ability to respond in real time.

💬

iMessage: native and trusted.

98% open rate. Read within 3 minutes. No spam folder. No carrier filtering. Shows your name and photo. This is where humans actually respond.

Architecture

The agent loop.

Your AI sends a message. The customer replies. Your AI responds. The whole loop runs in under 5 seconds.

Your AI Agent

Claude, GPT, Gemini, or custom

↓

Tuco REST API

POST /v1/messages

↓

iMessage

Delivered via real iPhone

↓

Customer

Reads and replies natively

↓

Reply Webhook

Fires in < 3 seconds

↺

Back to Your AI

Process reply, send next message

Use Cases

Agents that actually reach people.

💰

Sales Agent

Responds to inbound leads instantly. Qualifies based on your criteria. Books meetings on your calendar. Follows up if they go cold.

Lead fills out form → Agent texts in 30 sec → Qualifies → Books meeting → CRM updated

🛠️

Support Agent

Answers FAQs from your knowledge base. Handles repeat purchase questions. Escalates to human when confidence is low.

Customer asks about refill → Agent checks order history → Sends reorder link → Done

📅

Scheduling Agent

Handles appointment booking, reminders, and rescheduling. No back-and-forth email threads. Just quick text conversations.

Patient due for checkup → Agent texts → Patient picks time → Calendar updated

🚀

Onboarding Agent

Guides new customers through setup steps. Sends instructions, checks in on progress, answers questions along the way.

New signup → Welcome text → Setup guide → Check-in at 24h → Support if stuck

Code Example

The full agent loop in 20 lines.

Receive a webhook, call your AI, send the reply. That's it.

webhook-handler.ts

// 1. Receive reply via webhook
app.post('/webhook/reply', async (req, res) => {
  const { from, body } = req.body

  // 2. Call your AI model
  const aiReply = await claude.messages.create({
    model: 'claude-sonnet-4-20250514',
    system: 'Reply in under 160 chars. 
      Conversational. No emojis.',
    messages: [{ role: 'user', content: body }]
  })

  // 3. Send reply via Tuco
  await tuco.messages.send({
    to: from,
    body: aiReply.content[0].text
  })

  res.json({ ok: true })
})

Why Tuco

Not another messaging API.

Real iPhones, not VMs

Messages send from real Apple hardware. Your agent's messages show the sender's name and photo — not a random 5-digit shortcode. Trust is built in.

No "wall of text" problem

AI models love writing paragraphs. That's terrible for texting. Tuco's recommended system prompt enforces message-appropriate length — concise, natural, conversational.

Sub-3-second webhooks

Your agent needs to respond fast. Webhooks fire within 2-3 seconds of the customer's reply. The full AI loop completes before they put their phone down.

Human handoff built in

Set a confidence threshold. When your AI isn't sure, route to a human via Slack, your CRM, or a custom dashboard. Tuco handles the messaging — you handle the routing.

In Production

Already shipping.

Real companies are running AI agents on Tuco today. Here's what that looks like.

Case Study: Pharmaceutical Company

AI agent auto-replies to repeat purchase FAQs

Problem

Support team drowning in repeat questions — "when's my next refill?", "can I change my order?", "what's the dosage?"

Solution

AI agent receives customer iMessages via Tuco webhook, checks order history, and replies with personalized answers in under 5 seconds.

Handoff

When the AI's confidence drops below 80%, it routes to a human rep. Customer never knows the difference.

Result

70%+ of repeat questions handled autonomously. Support team focuses on complex cases. Response time dropped from hours to seconds.

FAQ

Questions. Answered.

Can I build a fully autonomous agent with Tuco?

Yes. Set up a webhook to receive replies, pipe them through your AI model (Claude, GPT, custom), and send the response back through Tuco's API. The loop runs 24/7 without human intervention. Most customers add a confidence threshold to escalate to a human when the AI isn't sure.

What about Apple's Terms of Service?

Tuco runs on real Apple hardware — dedicated iPhones and Macs, not emulators or VMs. Messages send through native iMessage infrastructure the same way you'd send from your personal device. We stay fully compliant.

How fast are webhooks for agent reply loops?

Webhooks fire within 2-3 seconds of the event. For an AI agent loop — customer replies, webhook fires, AI processes, Tuco sends reply — the total round-trip is typically under 5 seconds. Feels like texting a real person.

How do I prevent my agent from sending walls of text?

Tuco's system works with messages, not emails. Our recommended approach: include a system prompt that enforces message-appropriate length (under 160 characters for texts). Some customers also use our proprietary text refinement that compresses AI output into natural, concise messages.

Can my agent send images, PDFs, and voice notes?

Absolutely. Tuco supports images, videos (up to 25MB), PDFs, contact cards, and audio files. Your agent can send a product photo, a contract PDF, or a voice memo — all natively through iMessage.

What happens when my AI agent can't handle a question?

Build a confidence threshold into your agent logic. When the AI's confidence drops below your threshold, route the conversation to a human via Slack, your CRM, or a custom dashboard. Tuco handles the messaging layer — you handle the routing.

Give your agent a voice.

Your AI is ready. The API is ready. Connect them and start reaching customers where they actually respond.

Build Your Agent View API Docs

AI Agents Use Case Developer Docs Pricing Integrations