Skip to content

Active Graph

An event-sourced reactive graph runtime for long-running, auditable, agentic systems. Behaviors react to events, mutate the graph, emit more events. The event log is the source of truth — every run is replayable, forkable, and diff-able from its log.

You're already past pip install activegraph and you want to know where to start. Run the framework first, read second.

activegraph quickstart

That runs the bundled Diligence pack on fixtures (no API key, no configuration, under 30 seconds) and prints a memo. You'll see what the framework does before you read about how it does it. The command ends with a "what just happened" section pointing back here.

Start here

  • Quickstart — 10 minutes from install to a working custom behavior. Every example is a runnable file. Reading this is the canonical first thing.
  • Concepts: Failure model — read this second. The framework's stance on what counts as a recoverable failure governs how every error message reads and how every behavior should be written. Short on its own; load-bearing across everything else.

When something specific breaks

Every error message ends with a More: link to a page that explains the specific failure in detail — when it fires, what causes it, how to diagnose, how to fix. The error reference catalog is under Reference: Errors. You should rarely need to visit it directly; the error message that fired tells you which page.

When you're building something

  • Guides — concrete how-tos for writing behaviors, LLM behaviors, tools, pattern subscriptions, packs, and operating a run in production.
  • Concepts — the model's primitives (graph, events, behaviors, relations, patches, views, frames, policies, replay, forking) explained one at a time.
  • Cookbook — recurring patterns with copy-pasteable code.

Source and issues