Skip to content
Coro Docs

Concepts

This section explains how Coro behaves—layers, agents, workflows, memory, self-improvement, plugins, and the job harness—so you can configure tenants and author overlays with confidence.

Key topics

  • Intelligence layers — Base, tenant, and repo overlay merge rules.
  • Agents — The specialised roles that execute each phase.
  • Workflows — STANDARD, FAST, DEEP, and campaign pipelines.
  • Skills — On-demand knowledge invoked through the Skill tool.
  • Memory — Durable markdown knowledge and how it is read.
  • Self-improvement — propose_change and human-reviewed PRs.
  • Plugins — SCM, Tracker, and Executor extension points.
  • Deployment modes — Solo SQLite vs hybrid team control plane.
  • Job lifecycle — Phases, parking, webhooks, checkpoints.

Historical and specification-oriented pages—including architecture overviews synced from the framework repository—remain linked from the sidebar alongside the topics above.

Synced docs vs authored docs

Some concept pages carry a coro-docs-sync stamp (and often an upstream editUrl) because they are automatically synced from coro-developer-framework whenever docs/ changes on main.

  • Synced pages — Edit in the framework repo; the docs site ingest opens a PR in coro-web with the Markdown update.
  • Authoritative hand-written pages for this subsite — Live directly in coro-web (for example the overview pages marked without a sync banner).

Synced automation opens a pull request for human review; changes are merged only after normal review—not silently auto-merged without oversight.