Bring your AI.
Control your hardware.

Embedible is a hosted infrastructure layer for hobbyist hardware. It hands your devices to any MCP-compatible AI client (Claude Code, Cursor, etc.) over a secure MQTT-over-mTLS connection, plus gives you a browser-based provisioning flow, a wiring-diagram editor, and a live dashboard. You bring the AI — we handle the device interface.

Any client that speaks the Model Context Protocol over the Streamable HTTP transport. We test against Claude Code; the standard SDK clients work too. You generate an API key in the app, paste a snippet into your client's MCP config, and your devices appear as tools your AI can call.

Raspberry Pi Pico W, Pico 2 W, and ESP32 (WROOM-32, DOIT DevKit V1) are first-class today. ESP32-S3 should work as well. ESP8266 is not supported — its 80 KB SRAM cannot fit the AWS IoT TLS handshake.

Every device gets its own X.509 certificate, issued during provisioning and stored only on the device (the private key never sits on our servers after handoff). The AWS IoT policy attached to each cert is scoped to that device's own topic prefix, so a compromised device can only talk about itself.

No. Provisioning is point-and-click in the browser, and once a device is online you can ask your AI to write and push MicroPython code, read sensors, and update the dashboard. If you do know embedded, the wiring editor and pin-out tools are there for you.