Drift & Sync

Architecture models go stale as code changes. Someone refactors a service, adds a new endpoint, removes a dependency — and the diagram no longer matches reality. Scryer detects this automatically.

How drift detection works

Scryer monitors the source-mapped files in your project. It flags drift in two ways:

1. Changed files — a source-mapped file was modified since the last sync
2. New files — files appear in the project that aren’t covered by any node’s source map

When drift is detected, a sync bar appears at the bottom of the editor.

The sync bar

The sync bar shows:

• The connected agent’s name
• A list of potentially drifted nodes — click any name to navigate to it on the canvas
• A Sync button to start the sync process

Syncing

Click Sync to have Scryer spawn your connected AI agent with the MCP server attached. The agent:

1. Receives a list of potentially drifted nodes and their source patterns
2. Reads the changed files
3. Updates the model where the code has actually diverged
4. Model changes appear on your canvas in real-time

The canvas is locked during sync — you can watch the agent work but can’t make conflicting edits.

Canceling a sync

Click Cancel during a sync to roll back all changes the agent made. The model reverts to its state before the sync started.

Agent spawning

Scryer uses CLI spawning for sync:

• Claude Code — spawned via claude -p (uses your Anthropic subscription)
• Codex — spawned via codex exec (uses your OpenAI API key)

The MCP server is attached automatically. For Claude Code, the config is passed inline via --mcp-config. For Codex, it reads from the project’s .codex/config.toml.

Dismissing drift

If the drift notification is a false positive — files changed but the architecture didn’t actually diverge — click Dismiss to reset the baseline. Scryer won’t flag those changes again.

Keeping models fresh

For best results:

• Set source mappings on containers and components so Scryer knows which files to watch
• Run sync regularly after significant code changes
• Use broad glob patterns for containers (src/api/**) and narrow paths for operations (src/api/routes/users.ts:24-48)