About this book
The Living Guide to OpenClawis a new kind of technical reference. Rather than being written once and slowly going stale, it's continuously researched, verified, and refreshed on a weekly editorial cycle.
The editorial process
Every week, each chapter goes through a structured editorial cycle designed to keep the content accurate and trustworthy.
Research
Current sources are checked against each chapter's scope: official documentation, academic papers, community discussions, and release notes. Most weeks, most chapters require no changes — and that's a perfectly valid outcome.
Writing & revision
When research surfaces something worth incorporating, the relevant chapter is revised. New content is drafted, existing prose is refined, and the chapter's narrative is preserved.
Review & verification
Every revision is reviewed before publication. Sources are verified, claims are cross-checked, and the overall quality is assessed. Only approved changes make it to the published edition.
Transparent publishing
Every update is accompanied by edition notes: a plain-language explanation of what changed, why it changed, and the sources behind it. Nothing is updated silently.
Our principles
Accuracy over speed
We'd rather skip a weekly update than publish something unverified. If something can't be confirmed from authoritative sources, it doesn't make it in.
Clarity over comprehensiveness
This is a guide, not an encyclopaedia. We aim for the clearest possible explanation of each topic, not the most exhaustive.
Sourced and transparent
Every factual claim can be traced back to a source. Edition notes make it clear exactly what changed and why. Trust is earned through transparency.
Technical details about the editorial system
The editorial pipeline is powered by a team of specialised AI agents built with PydanticAI. A research agent gathers information from authoritative sources (official docs, academic databases, community forums). A writer agent produces drafts. A reviewer agent verifies claims and checks source quality. A publisher produces the final editions in multiple formats.
All content and metadata is stored in PostgreSQL with append-only versioning. The web reader is built with Next.js and Fumadocs. PDF and EPUB editions are generated with Typst and Pandoc respectively.
The entire system runs on a single Hetzner server, orchestrated with Docker Compose, with weekly cron-triggered editorial cycles.