Skip to main content

ctxl — Context Engineering CLI for AI Agents

Reduce token waste. Prevent hallucination. Zero AI used.

ctxl (pronounced "contextual") is a developer CLI tool that helps you manage the context window of AI coding agents (GitHub Copilot, Cursor, Claude, etc.) by generating compressed codebase skeletons, task-focused instructions, and session checkpoints — all using deterministic parsing, not AI.

The Problem

AI coding agents (Copilot, Cursor, etc.) read your entire codebase to build context. A 3,000-line PySpark file burns ~750 tokens every time the agent references it. Over a session, your context window fills with noise, leading to:

  • 🔴 Hallucination — the model starts making things up
  • 🔴 Token waste — you pay for irrelevant context
  • 🔴 Lost focus — the agent forgets your actual task

The Solution

ctxl gives you three commands that prepare your environment before the AI agent reads it:

ctxl map          # Generate a compressed codebase skeleton (~95% token reduction)
ctxl init         # Generate task-focused Copilot instructions
ctxl checkpoint   # Save session state for safe /clear workflows

Zero AI models. Zero API calls. Zero tokens burned by this tool.

Installation

pip install ctxl-cli

Quick Start

ctxl map — Codebase Skeleton

Generate a compressed structural map of your codebase with line numbers:

ctxl map                      # Map current directory
ctxl map ./src                # Map a specific directory
ctxl map -e .py               # Only Python files
ctxl map -o codebase.md       # Save to file
ctxl map --clipboard          # Copy to clipboard for pasting into AI chat

Before (raw file, ~750 tokens):

class DataPipeline:
    def __init__(self, spark, config):
        self.spark = spark
        self.config = config
        self.source_path = config.get("source_path", "/data/raw")
        # ... 60 more lines of implementation
    
    def clean_data(self, df, drop_nulls=True):
        string_cols = [f.name for f in df.schema.fields ...]
        # ... 20 more lines

After (ctxl map output, ~50 tokens):

L22: class DataPipeline:
    L25: def __init__(self, spark: SparkSession, config: Dict)
    L33: def load_data(self, table_name: str, filters: Optional[Dict] = None) -> DataFrame
    L42: def clean_data(self, df: DataFrame, drop_nulls: bool = True) -> DataFrame
    L52: def transform(self, df: DataFrame, rules: List[Dict]) -> DataFrame
    L66: def validate(self, df: DataFrame) -> bool
    L75: def save(self, df: DataFrame, partition_cols: List[str] = None) -> str

Line numbers (L22:, L42:) let the AI agent navigate directly to the right location.

ctxl init — Copilot Instructions

Generate a .github/copilot-instructions.md file that GitHub Copilot reads natively:

ctxl init "Fix the data pipeline ETL bug"
ctxl init "Add authentication" -f auth.py -f models.py
ctxl init "Refactor tests" --no-map

ctxl checkpoint — Session State

Save your progress before running /clear in Copilot Chat:

ctxl checkpoint save \
    -t "Fix ETL pipeline" \
    --done "Found the bug in clean_data()" \
    --state "Pipeline runs but output has wrong column order" \
    --next "Fix column ordering in transform()" \
    --file "data_pipeline.py"

ctxl checkpoint list          # List all checkpoints
ctxl checkpoint show          # Show latest checkpoint

Supported Languages

ctxl map uses Tree-sitter for parsing and supports:

Language Extensions
Python .py
JavaScript .js, .jsx
TypeScript .ts, .tsx
Java .java

More languages can be added easily via Tree-sitter grammars.

How It Works

Your Codebase (10,000+ tokens)
        │
        ▼
   Tree-sitter Parser (deterministic, local, free)
        │
        ▼
   AST → Extract signatures + line numbers
        │
        ▼
   Compressed Skeleton (~500 tokens)
        │
        ▼
   AI Agent reads skeleton instead of raw code
        │
        ▼
   90-95% fewer tokens burned 🎉

License

MIT

Metadata

Release files for ctxl-cli 0.4.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for ctxl-cli 0.4.0
File Size Uploaded Provenance Actions
ctxl_cli-0.4.0.tar.gz 19.3 kB With provenance Details for ctxl_cli-0.4.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for ctxl-cli 0.4.0
File Interpreter ABI Platform Actions
ctxl_cli-0.4.0-py3-none-any.whl Python 3 none any Details for ctxl_cli-0.4.0-py3-none-any.whl

Total release size: 39.3 kB

Release files / ctxl_cli-0.4.0.tar.gz

Download URL ctxl_cli-0.4.0.tar.gz
Size 19.3 kB
Tags Source
SHA-256 checksum
How to use checksums
e1396fef38f2362747ab92c5f94295b2503714955625e29c68bd9a780e4d7c82
BLAKE2b-256 checksum
How to use checksums
f2b7f3d5c21551c5a6d38d151f074016842fd560f741b85eada253ab23c437a3
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

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.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 15, 2026.

Transparency log

Release files / ctxl_cli-0.4.0-py3-none-any.whl

Download URL ctxl_cli-0.4.0-py3-none-any.whl
Size 20.0 kB
Tags Python 3
SHA-256 checksum
How to use checksums
06b52ad2f4c13f86a563e7686118f29867d2aa7372844dde6952caf090808ef1
BLAKE2b-256 checksum
How to use checksums
7c3b7c4dbd5b1146cd6d046f27cf98198ad011b899137944964873b1bd5b06a4
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

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.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 15, 2026.

Transparency log

Security for ctxl-cli 0.4.0

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.

Full provenance

Both files in this release have been uploaded with provenance metadata.

Select from 2 files

Attestation for ctxl_cli-0.4.0.tar.gz

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 15, 2026.

Transparency log

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.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page