Core
Project-level documentation — the why, what, and where-next of AgentCanvas. Start here if you're onboarding or returning after a break.
Main pages
Blueprint
Problem statement, 9 capabilities, design principles, core components, differentiation, open questions, status.
Architecture
System diagram, data flow, flat workspace, execution engines, component system, state containers, tech stack.
Codebase Map
Per-package file walkthrough — every file in agentcanvas/ grouped by package with role descriptions.
Glossary
Load-bearing terms, grouped by 10 subsystems — searchable / filterable in v2.
Roadmap
Open work — maintenance TODO, features (F·), env integrations (E·), method ports (M·), planned, deferred.
Decisions
Architecture decision records grouped into 8 fields. Context · Decision · Alternatives · Rationale.
Going deeper
Two further layers sit beneath core: capabilities (the nine things AgentCanvas does, with implementation walkthroughs) and design-docs (internal subsystem specs).
Capabilities →
The nine capabilities from blueprint — each with a dedicated page on mechanism, key files, and current status. Linked prev/next so you can read straight through.
Design Docs →
22 deep-dive references in five groups — components (Canvas System / Nodesets / Env Panels / Env Lifecycle), graph (Graph System / Graph Executor / Wire Types / State Containers / Loop Control / Batch Eval), operations (Server Mode / LLM Config / Execution Logs / Reload & Code Freshness), surfaces (Coding-Agent Backend Surface / Search-Operator API), and usage.
Suggested first reads in design-docs: Canvas System · Graph System · Graph Executor · State Containers.