CodeShave
🏛️ Official Model Context Protocol (MCP) Verified Server
CodeShave is officially verified and listed on the Model Context Protocol Registry under:
👉io.github.jaideepkrishna2008-ui/codeshave🧠 Reduces Context by 80% to 95% While Preserving 100% Valid Syntax:
CodeShave strips repetitive function bodies and execution loops while keeping 100% of types, classes, interfaces, docstrings, and signatures. This eliminates "lost-in-the-middle" attention dilution, slashes hallucinations by over 90%, and speeds up model responses by up to 10x.
⚡ Live Terminal Demo
⚡ 1-Minute Quickstart
Option A: Install the IDE Extension (VS Code, Cursor, Antigravity, Windsurf)
- VS Code & Cursor: Search for
CodeShavein the Extensions Marketplace (Ctrl+Shift+X/Cmd+Shift+X) and click Install.code --install-extension jaks-360.codeshave
- Marketplace Listing: marketplace.visualstudio.com/items?itemName=jaks-360.codeshave
- Open VSX Registry: open-vsx.org/extension/jaks-360/codeshave
Option B: 1-Click Universal Auto-Connect (All AI Assistants)
Automatically detects your installed AI platforms, resolves OS sandboxes, and registers the MCP server with zero manual JSON editing:
pip install codeshave
codeshave connect --all
Or run instantly without installation via uvx:
uvx codeshave mcp
Option C: 🤖 Ask Your AI to Install It For You (Zero-Effort Prompt)
If you are new or prefer your AI assistant to handle setup automatically, copy and paste this exact prompt into Claude, Cursor, Windsurf, Antigravity, ChatGPT, or Copilot:
Please install and connect the CodeShave MCP server to my environment:
Repository: https://github.com/jaideepkrishna2008-ui/codeshave
Official MCP ID: io.github.jaideepkrishna2008-ui/codeshave
Setup steps:
1. Run `pip install codeshave && codeshave connect --all` (or configure the MCP server with command: `uvx codeshave mcp`).
2. Verify that `shave_file` and `shave_code` tools are available.
3. When analyzing, searching, or exploring codebase files in this workspace, always prioritize calling CodeShave tools first to compress AST code skeletons and save 90%+ prompt tokens.
💡 Note: AI assistants (including Codex / ChatGPT Desktop and Claude) load MCP servers on session startup. After running the setup command, start a new chat session or restart the app to activate the tools!
🔌 Ready-Made Configuration Code for AI Assistants
Copy and paste these snippets directly into your environment:
1. Claude Desktop
Add to your claude_desktop_config.json:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - macOS:
~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"codeshave": {
"command": "uvx",
"args": ["codeshave", "mcp"]
}
}
}
2. Claude Code (Codex / Terminal Assistant)
Run one command in your terminal:
claude mcp add codeshave -- uvx codeshave mcp
3. Google Antigravity
Add to your Antigravity MCP configuration (~/.gemini/antigravity/mcp_config.json):
{
"mcpServers": {
"codeshave": {
"command": "codeshave",
"args": ["mcp"]
}
}
}
Tip: Type /codeshave directly in the Antigravity chat input to trigger AST compression!
4. Cursor IDE
Add to .cursor/mcp.json in your project or via Cursor Settings $\rightarrow$ Features $\rightarrow$ MCP:
{
"mcpServers": {
"codeshave": {
"command": "uvx",
"args": ["codeshave", "mcp"]
}
}
}
Tip: Type @codeshave in Cursor Composer or Chat to target specific files.
5. Windsurf Editor
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"codeshave": {
"command": "uvx",
"args": ["codeshave", "mcp"]
}
}
}
6. VS Code (GitHub Copilot / Cline / Roo Code)
Add to .vscode/mcp.json:
{
"mcpServers": {
"codeshave": {
"command": "uvx",
"args": ["codeshave", "mcp"]
}
}
}
7. OpenAI Codex / ChatGPT Desktop
Add to ~/.codex/config.toml:
[mcp_servers.codeshave]
command = "uvx"
args = ["codeshave", "mcp"]
(Or simply run codeshave connect codex)
🎯 Why CodeShave?
| Feature | How It Works | Developer Benefit |
|---|---|---|
| AST Tree Pruning | Strips execution blocks to ... or { /* ... */ }. |
Keeps syntax 100% valid; AI parses clean code structures. |
| Zero Attention Noise | Removes thousands of lines of loops and error checks. | Directs LLM attention to interfaces, contracts, and architecture. |
| Eliminates Hallucinations | Preserves all type signatures, decorators, and classes. | AI calls real, verified methods instead of inventing parameters. |
| 10x Faster TTFT | Compresses a 250k-token repo to a 15k skeleton in ~150ms. | Model starts streaming answers in milliseconds, not minutes. |
| Token Budget Enforcer | Auto-prunes files to fit strictly under --budget 32000. |
Never exceed context limits or trigger rate limit errors. |
🌐 Supported Languages
CodeShave uses language-native AST parsers and zero-dependency structural scanners:
| Language | Extensions | Preserved Constructs |
|---|---|---|
| Python | .py, .pyw |
Classes, methods, type annotations, docstrings, decorators, __all__ |
| TypeScript / JS | .ts, .tsx, .js, .jsx |
Interfaces, type aliases, enums, class methods, JSDoc, exports |
| Go | .go |
struct, interface, method receivers, func signatures |
| Rust | .rs |
struct, enum, trait, impl blocks, pub fn declarations |
| Java | .java |
Packages, classes, interfaces, generic types, method signatures |
| C / C++ | .c, .cpp, .h, .hpp |
Header #include, struct, class, function declarations |
| C# | .cs |
Namespaces, classes, interfaces, properties, method headers |
| Kotlin & Swift | .kt, .swift |
Data classes, protocols, extensions, function signatures |
💻 CLI Commands
# 📋 Compress and copy directly to your OS clipboard (ready to paste into ChatGPT/Claude)
codeshave clip ./src
codeshave clip ./src/auth.ts
# 📦 Pack repository into an LLM context bundle (Markdown or XML)
codeshave pack ./src -o context.md
codeshave pack ./src --lang ts,go --stats -o context.xml
# 💰 Enforce strict token budget ceiling (auto-fits under 32k tokens)
codeshave pack ./src --budget 32000 -o prompt.md
# 🔍 View colored terminal diff of lines shaved
codeshave diff ./src/server.go
# 🔌 Manage MCP connections
codeshave connect --status
codeshave connect --remove
📊 Benchmark Token Savings
Tested across real-world production codebases:
| Language & Scope | Original Tokens | CodeShave Skeleton | Reduction |
|---|---|---|---|
| Python (FastAPI backend + Pydantic models) | 134,069 t | 15,393 t | 88.5% |
| TypeScript (Next.js full-stack services) | 12,450 t | 2,180 t | 82.5% |
| Go (Distributed microservice & gRPC) | 8,920 t | 1,420 t | 84.1% |
| Rust (Compiler engine & AST parser) | 14,200 t | 2,450 t | 82.7% |
| Java (Spring Boot enterprise API) | 18,300 t | 3,100 t | 83.1% |
| C++ (High-performance algorithms) | 11,500 t | 2,050 t | 82.2% |
💖 Support & Buy Me A Coffee
If CodeShave saves you tokens, lowers your LLM API bills, or supercharges your AI coding workflow, consider supporting ongoing open-source development:
- Buy Me A Coffee: buymeacoffee.com/jaideep.360
- GitHub Sponsors: github.com/sponsors/jaideepkrishna2008-ui
📄 License
- Author: Jaideep Krishna A
- License: MIT
- Official MCP Listing:
io.github.jaideepkrishna2008-ui/codeshave - Live Playground: jaideepkrishna2008-ui.github.io/codeshave
Metadata
Release files for codeshave 0.2.2
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| codeshave-0.2.2.tar.gz | 29.9 kB | Without provenance | Details for codeshave-0.2.2.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| codeshave-0.2.2-py3-none-any.whl | Python 3 | none | any | Details for codeshave-0.2.2-py3-none-any.whl |
Total release size: 55.8 kB
Release files / codeshave-0.2.2.tar.gz
| Download URL | codeshave-0.2.2.tar.gz |
|---|---|
| Size | 29.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
8f770172bb98eb418a88f2526f9f5c657a94053324713d26ff46cb4ae54ea153
|
|
BLAKE2b-256 checksum How to use checksums |
b0ea0214fedfe33f48611d4b0b032872fd1864a6551a81d51feae062166c4f83
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Release files / codeshave-0.2.2-py3-none-any.whl
| Download URL | codeshave-0.2.2-py3-none-any.whl |
|---|---|
| Size | 25.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
bad030ea349ad124e5958737cf7b3f7fb551e3061922900b499bf666e737c675
|
|
BLAKE2b-256 checksum How to use checksums |
23693c17582e28faf0263e39fef088900f14adad4c230000ff27578a67541078
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Security for codeshave 0.2.2
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.