Skip to content

scryer

Visual architecture models that AI agents build from.

Scryer is a visual architecture tool for working with AI coding assistants. You draw C4 diagrams in a drag-and-drop editor — AI agents read, modify, and implement from the same model through MCP.

The core idea: natural language is a lossy way to tell an AI what to build. A visual model forces the kind of precision that prose can’t. You see the same architecture the agent sees — misalignment gets caught before code gets written.

  1. Model your architecture — drag nodes onto the canvas, draw relationships, drill into detail
  2. Connect your AI tool — Claude Code, Codex, or any MCP-compatible agent
  3. The agent reads your model — it sees exactly what you see: systems, containers, components, data flows
  4. You iterate visually — the agent proposes changes, you see them on the canvas in real-time
  5. Implementation follows the modelget_task feeds the agent one piece of work at a time, ordered by dependencies

Scryer uses a C4-based hierarchy. Each level zooms in further:

LevelWhat it represents
PersonUsers and actors
SystemSoftware systems
ContainerApps, services, data stores
  ↳ ComponentLogical modules
    ↳ OperationFunctions & handlers
    ↳ ProcessMulti-step workflows
    ↳ ModelData structures

Person and System are top-level. Select a system and click the expand button to see its containers. Expand a container to see its components. Components contain the code-level detail: operations (functions), processes (workflows), and models (data structures).

  • Drag-and-drop editor with multiple node shapes, auto-layout, and zoom controls
  • Behavioral flows — model user journeys and data pipelines with branching
  • Contracts — tell AI agents exactly how to implement each piece, with pass/fail tracking
  • Status tracking — four statuses: Proposed Implemented Verified Vagrant
  • Source mapping — link architecture nodes to files in your codebase
  • Drift detection — Scryer watches your code and flags when the architecture model goes stale
  • AI advisor — optional LLM review that flags structural issues in your diagrams
  • Themes — customize every color in the app