CLI Reference

Sync Operations

Spiral keeps your local IDE files, the central knowledge graph, and every connected agent in sync. Synchronization can happen manually via the CLI or automatically via MCP tools.

Status & Diff

Check the current state before syncing.

$ spiral status
Health Score: 98/100
Active locks: 0
Pending proposals: 2
Unresolved conflicts: 1
$ spiral diff
+ [stack] Added Tailwind CSS v4
- [stack] Removed Bootstrap 4
~ [decision] Updated auth strategy

Push & Pull

Synchronize your local state with the Spiral server.

$ spiral pull
Fetching latest...
Updated 3 files (CONTEXT.md, .cursor/rules, CLAUDE.md)
$ spiral push -m "Added payment module"
Detecting changes...
Pushed context update. IDE files regenerated.
$ spiral sync
Bidirectional sync...
Local → server: 1 change. Server → local: 2 files updated.

Auto-sync via MCP

When working in an MCP-connected IDE, your AI agent can synchronize automatically using two built-in tools:

complete_feature

Called when the agent finishes a feature. Creates a provenance record, extracts knowledge from changed files, checks for conflicts, and regenerates all IDE projections.

checkpoint

A lightweight save point. Snapshots the current state without running full extraction — useful for work-in-progress saves.

Drift detection

Spiral automatically detects when your project files and your CONTEXT.md have drifted apart — for example, when files are renamed or deleted but the context still references them. Running spiral sync or the MCP sync_context tool resolves drift by regenerating projections from the current knowledge graph.