Blog

Building in the open

Updates, tutorials, architecture deep dives, and lessons learned building AI agent tooling from a Raspberry Pi.

Introducing ClawTools - open-source tool connectors for AI agents

Why we built ClawTools, what problem it solves, and how it fits into the PicoClaw/OpenClaw ecosystem. The story of going from "I want a morning briefing on my Pi" to an open tool registry.

→

Setting up your first morning briefing with Gmail + Google Calendar

Step-by-step: from a fresh Raspberry Pi to receiving a daily AI-generated briefing on Telegram every morning at 7am.

→

How MCP tool calls work - and why they're the right abstraction for AI agents

A technical walkthrough of the Model Context Protocol, how tool call schemas are defined, and why ClawTools uses a dual MCP + HTTP interface.

→

The OAuth2 reality check: Google vs Microsoft, verified apps, and what it actually takes

Everything we learned navigating Google's OAuth verification process and Microsoft's Partner Network requirements. The real checklist, real timelines, real costs.

→

Building a Go microservice that speaks both MCP and REST

How ClawTools exposes the same tool definitions over both stdio MCP transport and HTTP REST with a single unified codebase in Go.

→

More posts coming as we build. Follow the repo to stay updated.