toonify-mcp

Toonify MCP
==============

What it is
- Open-source Claude Code plugin and MCP server for local context compression.
- Compresses large tool output before it enters the context window, reducing token cost without changing the Claude Code workflow.

What it compresses
- Structured data: JSON, YAML, API responses
- Debug-heavy output: test failures, stack traces, compiler diagnostics, repetitive lint/build output

What stays the same
- Claude Code workflow in plugin mode
- Local-first processing model
- Safe fallback when estimated savings are not worth it

Best for
- Workflows with frequent large structured data or API response reads into Claude Code
- Debugging sessions with long test failures, stack traces, or repetitive build output

Limitations
- Short prose
- Very small files
- Tasks where preserving exact original formatting matters more than saving context

How to install
1. Clone the repo and build it locally
2. Install `toonify-mcp` from the local checkout
3. Run `toonify-mcp setup` for plugin mode
4. Run `toonify-mcp setup mcp` only if you need MCP server mode

How to verify
- `toonify-mcp setup`
- `toonify-mcp doctor`
- `toonify-mcp status` for the latest optimized or skipped outcome
- `claude mcp list` when MCP mode is enabled

Public proof
- Repo: https://github.com/PCIRCLE-AI/toonify-mcp
- Benchmarks: https://toonify.pcircle.ai/benchmarks.html
- Benchmark suite source: https://github.com/PCIRCLE-AI/toonify-mcp/blob/main/tests/benchmarks/quick-stats.test.ts

Current structured-data benchmark snapshot
- 12 fixtures
- 48.1% average savings
- 24.5-66.3% observed range

Primary pages
- https://toonify.pcircle.ai/
- https://toonify.pcircle.ai/index-zh.html
- https://toonify.pcircle.ai/benchmarks.html
- https://toonify.pcircle.ai/benchmarks-zh.html
- https://toonify.pcircle.ai/privacy.html
- https://toonify.pcircle.ai/privacy-zh.html
- https://toonify.pcircle.ai/terms.html
- https://toonify.pcircle.ai/terms-zh.html