Your codebase.
Compressed.
Ready to hand off.
Stop wasting tokens. TokenCap snapshots your workspace into structured, AI-ready prompts with files, imports, Git context, and invariants.
TokenCap gives you one command for the entire handoff story.
Deterministic AST compression, verified call graphs, static taint tracking, and auto-wired agent pointers all local, zero cloud required.
One command. Complete context handoff.
Parses your repository AST, extracts Git diffs, scrubs secrets, and formats compressed LLM-ready markdown in 82ms.
Token Savings12.2× average
Execution Time82ms
Output Files3 artifacts
[tokencap] Indexing workspace tree (148 files)...
✓ Parsed AST graph in 82ms
✓ Ranked 24 relevant context files (Git staged & diff priority)
✓ Scrubbed 4 secrets (OpenAI API key, AWS secret, GitHub PAT)
→ Created TOKENCAP.md38.2 KB · 9,450 tokens
→ Created TOKENCAP_GRAPH.md14.1 KB · 3,520 tokens
→ Created TOKENCAP_MEMORY.md6.8 KB · 1,700 tokens
Total context packaged:14,670 tokens (12.2× reduction from 820K)
Three simple steps. No cloud required.
From editor keystroke to structured AI context in sub-80ms — running 100% locally on your machine.
TRIGGER01
Intercept on save
Hit Ctrl+S in your editor. TokenCap's background daemon instantly dirty-checks changed files in memory without scanning disk.
tokencap watch --debounce=30ms< 1ms trigger
PROCESS02
Compile AST delta
The incremental parser resolves AST call-graphs, filters out dead code, redacts secrets, and compiles a fresh TOKENCAP.md.
→ TOKENCAP.md (38.2 KB)82ms compile
HANDOFF03
Auto-wire AI host
The local stdio MCP server exposes 11 tools directly to Claude, Cursor, Windsurf, or Cline with zero network egress.
stdio://127.0.0.1 (11 tools)0 cloud uploads
Enjoying TokenCap?Leave a review →
Ready to sync?
Stop copy-pasting directories. Automate your LLM context packaging in one command and build 10x faster.
