CodeShave
High-speed, deterministic AST code context compressor for AI coding assistants and local LLMs.
Reduces prompt tokens by 60% to 80% while preserving valid syntax, type signatures, and class hierarchies.
⚡ Terminal Demo
🎯 Why CodeShave?
When developers feed large directories to LLMs or AI agents, context windows quickly fill up, latency spikes, and API token bills mount.
Traditional approaches either:
- Truncate files blindly, breaking syntax trees and references.
- Rely on costly, slow LLMs to summarize code before prompting.
CodeShave uses Python's native Abstract Syntax Tree (AST) engine to deterministically transform code into clean interface skeletons:
- Strips implementation logic inside functions and replaces with
.... - Preserves all class structures, inheritance, and attributes.
- Retains parameter type annotations, defaults, and return types.
- Maintains decorators, docstrings, and module-level exports.
- Formats context blocks into Markdown, Claude-optimized XML, or JSON.
🏗️ Architecture
🚀 Quickstart
Installation
Install directly via pip from GitHub:
pip install git+https://github.com/jaideepkrishna2008-ui/codeshave.git
Or clone for local development:
git clone https://github.com/jaideepkrishna2008-ui/codeshave.git
cd codeshave
pip install -e .
⚡ CLI Commands
# 📋 One-Click Clipboard Copy: Compress and copy directly to clipboard
codeshave clip ./src
codeshave clip ./src/models.py
# -> Copied to clipboard! Ready to paste directly into ChatGPT, Claude, or Cursor!
# 📦 Pack directory into an LLM context bundle
codeshave pack ./src -o llm_context.md
# 🏷️ Generate Claude-optimized XML bundle with live savings statistics
codeshave pack ./src --format xml --stats -o llm_context.xml
# 🔍 Inspect a single file's compressed skeleton
codeshave inspect ./src/services/billing.py
🔌 Model Context Protocol (MCP) Integration
CodeShave runs as a turnkey MCP Server over standard I/O, allowing Claude Desktop, Cursor, and Windsurf to automatically compress files on the fly before reasoning:
Add to your claude_desktop_config.json:
{
"mcpServers": {
"codeshave": {
"command": "codeshave",
"args": ["mcp"]
}
}
}
Python API
from codeshave import compress_source, compute_savings
raw_code = open("service.py").read()
skeleton = compress_source(raw_code, preserve_docstrings=True)
report = compute_savings(raw_code, skeleton)
print(skeleton)
print(report.summary_line())
📊 Token Savings Benchmark
| Source Type | Original Tokens | CodeShave Skeleton | Reduction | Syntax Status |
|---|---|---|---|---|
| Data Models | 3,460 t | 920 t | 73.4% | ✅ Valid AST |
| API Endpoints | 4,210 t | 780 t | 81.5% | ✅ Valid AST |
| Business Logic | 6,840 t | 1,120 t | 83.6% | ✅ Valid AST |
| Utility Scripts | 1,000 t | 400 t | 60.0% | ✅ Valid AST |
💖 Sponsorship & Backing
If CodeShave saves you tokens, reduces your API bills, or accelerates your agent workflow, consider supporting ongoing open-source development:
- Buy Me A Coffee: https://buymeacoffee.com/jaideep.360
- GitHub Sponsors: https://github.com/sponsors/jaideepkrishna2008-ui
📄 License
Created by Jaideep Krishna A (@jaideepkrishna2008-ui). Licensed under the Apache License 2.0 or MIT equivalent open-source terms.
Metadata
Release files for codeshave 0.1.0
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-0.1.0.tar.gz | 15.6 kB | Without provenance | Details for codeshave-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| codeshave-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for codeshave-0.1.0-py3-none-any.whl |
Total release size: 30.5 kB
Release files / codeshave-0.1.0.tar.gz
| Download URL | codeshave-0.1.0.tar.gz |
|---|---|
| Size | 15.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
504257391c69103ab57ec77de1064408b22fe8c19e5a608ef57bbb839aeaf6aa
|
|
BLAKE2b-256 checksum How to use checksums |
3826466d4ba016c85fb49bb591032392eac40bc37f989a57cd2a8dea6d5fcb94
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.11
|
Release files / codeshave-0.1.0-py3-none-any.whl
| Download URL | codeshave-0.1.0-py3-none-any.whl |
|---|---|
| Size | 14.9 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
07125c730c2f4c927c5012aa8b4ba102c887f8932b60eb75694c33c6e73ff4fa
|
|
BLAKE2b-256 checksum How to use checksums |
0ca9dc5edf9711d8a440da0541485fd54404263fa91534fea7d99d140716102a
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.11
|
Security for codeshave 0.1.0
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.