Skip to content

Concepts

Background reading for the pieces that make OpenClaw tick.

Pages

Reading list (in the upstream repo)

The repo's docs/concepts/ directory covers more topics not duplicated here:

  • agent.md, agent-loop.md, agent-workspace.md — the agent runtime.
  • architecture.md — the canonical architecture reference.
  • compaction.md, context.md, memory.md — context management.
  • messages.md, markdown-formatting.md — message handling.
  • model-failover.md, model-providers.md, models.md — model layer.
  • multi-agent.md — multi-agent routing.
  • oauth.md — provider OAuth flows.
  • presence.md, typing-indicators.md — presence model.
  • queue.md, retry.md, streaming.md — message lifecycle.
  • session*.md — session lifecycle, pruning, the session tool.
  • system-prompt.md — system prompt construction.
  • timezone.md, typebox.md, usage-tracking.md — supporting concepts.

These read well in the source repo's docs/concepts/ directory; this site curates the most important entry points.