Control complex AI agents via iMessage and SMS with Poke
The Interaction Company launches Poke, a $300 million AI agent platform that automates routines and manages apps through standard text messaging.
The Interaction Company of California launched Poke to the public, moving their AI agent platform out of an email-focused beta into standard text messaging applications. The system allows users to execute complex tasks and manage daily routines directly through iMessage, SMS, Telegram, and WhatsApp. For developers building consumer AI, the zero-install messaging interface bypasses the standard application friction layer.
Architecture and State Management
Poke uses a decoupled architecture to manage the user experience and the technical execution separately. A user-facing Personality Agent handles conversational tone, state management, and interface constraints. It routes requests to an Execution Agent that manages external API calls, web browsing, and data processing.
This multi-agent system ensures the user is isolated from intermediate processing steps. The Execution Agent passes structured results back to the Personality Agent. The system then formats the output for the specific text and media constraints of an SMS or iMessage thread.
Proactive Alerts and Capabilities
Most conversational interfaces require user prompting. Poke operates continuously in the background, monitoring connected accounts to surface proactive alerts. The system connects to Gmail and Google Calendar to track meeting changes and flight schedules. It also supports hardware integrations for Philips Hue and Sonos, alongside data pipelines from Strava, Fitbit, and Notion.
Users automate these systems using Recipes, which are agent skills defined via natural language. Poke parses the intent and maps it to the corresponding third-party API endpoints. The underlying platform maintains SOC 2 and CASA Tier 2 compliance for handling user authentication tokens and personal data.
Developer Tooling and Capital Structure
The platform includes an npx tool for engineers to write custom execution logic. Standard developer integrations cover GitHub, Vercel, Supabase, and Cursor. This allows developers to monitor repository changes or deployment statuses directly via text message.
During its summer 2025 closed beta, Poke processed over 200,000 monthly messages across 6,000 users. The current public release targets select markets where WhatsApp and Telegram dominate, including Brazil, with availability dictated by regional regulatory shifts.
Concurrent with the launch, the 10-person Palo Alto startup announced a $300 million post-money valuation. The company secured $10 million in new capital, bringing total funding to $25 million following a seed round led by General Catalyst.
| Funding Stage | Amount | Notable Investors | Valuation |
|---|---|---|---|
| Seed Round | $15 Million | General Catalyst | Undisclosed |
| New Funding | $10 Million | Spark Capital, Village Global | $300 Million |
Additional individual backers include the founders of Stripe, Vercel CEO Guillermo Rauch, and OpenAI’s Joanne Jang.
If you are building consumer AI applications, evaluate the delivery mechanism alongside the model capabilities. Forcing users to download a dedicated app or authenticate a web client introduces immediate drop-off. Building stateful AI agents directly into the communication channels users already check multiple times a day reduces activation energy to near zero.
Get Insanely Good at AI
The book for developers who want to understand how AI actually works. LLMs, prompt engineering, RAG, AI agents, and production systems.
Keep Reading
How to Build Advanced AI Agents with OpenClaw v2026
Learn to master OpenClaw v2026.3.22 by configuring reasoning files, integrating ClawHub skills, and deploying secure agent sandboxes.
Claude Sheets, Whiteboards and Meetings: Anthropic builds its own office suite
On November 10 Anthropic launched Claude Sheets and Whiteboards in beta, made Claude Data generally available, added interactive connectors from Adobe and Atlassian, and began a Claude Meetings beta that puts Claude into Google Meet calls.
DeepMind's AlphaProof Code proved Signal's security on the compiled binary: 6.5M lines of Lean, no human-written proofs
Google DeepMind's AlphaProof Code used AI agents to prove forward secrecy, post-compromise security and authentication for libsignal's AArch64 machine code in four weeks. Here is what was proved, what was assumed, and what the agents got wrong along the way.
SSHole: agent swarms found a ten-year-old OpenSSH login bypass, and attackers were using it within two days
CVE-2026-123158 lets a 257-byte password log in to accounts that have no password on musl-based Linux such as Alpine. A researcher found it by having AI agents fire about 17,600 odd logins at differently built sshd binaries. OpenSSH 10.6p2 fixes it.
Apple Moves to Lock Down Full Disk Access, Citing AI Agent Risks
Apple announced on October 2 that Full Disk Access will require much more explicit user consent going forward, citing AI agents specifically: 'as AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially.'