⚡ The Local Code Graph Engine for AI Coding Agents
Stop burning 90% of your context window on repo scans.
Canvizo maps your entire codebase into an instant semantic graph. Feed AI agents like Claude Code, Cursor, and Windsurf exact architectural subgraphs—completely local, air-gapped, and blazing fast.
✔ You're on the early access list! We will reach out shortly.
Zero telemetry. Your code stays 100% local on your machine.
90% Token Reduction
Replaces brute-force file scanning with precise bounded graph context. Stop burning budget on irrelevant files.
Raw Files Scan:
128,000 tokens
Canvizo Subgraph:
8,200 tokens
Sub-80ms Graph Queries
In-process Tree-sitter & AST parsing instantly constructs call graphs and dependency resolution. Zero cloud latency.
Query Speed
~42ms avg
Air-Gapped & Local
100% zero-egress architecture. Your source code and graph indexes never leave your local workspace.
Storage: .canvizo/index.db
$
npx @canvizo/mcp start --repo ./
i [Canvizo] Initializing local Tree-sitter graph engine...
✔ Mapped 1,420 files in 420ms | 0 external packets sent
▶ MCP Server listening on stdio (Ready for Claude Code / Cursor)