Entroly

RAG & Memory 💻 Python ⚖️ Apache-2.0 🟢 Actively maintained
443 stars

Entroly is an open-source, local-first AI token-efficiency and Context Assurance layer that reduces avoidable AI token usage and provider-bound context by selecting high-value evidence under explicit token budgets, compressing context recoverably, and emitting auditable receipts. It works through proxy, MCP, plugin, wrapper, and SDK paths with popular AI coding tools and SDKs, without replacing your model or agent architecture. It is for developers using AI coding assistants or building AI applications who want to cut token costs while keeping critical evidence recoverable and verifiable.

Entroly demo
🎞️ Demo from the project README

✨ Key features

  • Budgeted evidence selection under explicit token budgets
  • Recoverable context compression with byte-exact recovery
  • Content-addressed evidence recovery and auditable receipts
  • Works via proxy, MCP, plugin, wrapper, and SDK paths
  • Integrates with Claude Code, Codex, Copilot, Cursor, Aider, and more
  • Live token savings metrics and privacy-safe telemetry

🎯 Use cases

  • Reduce token usage and costs for AI coding assistants like Claude Code or Cursor
  • Compress long agent sessions to avoid context limits without losing data
  • Selectively include only relevant code fragments in prompts to improve model focus
  • Provide auditable receipts for AI context decisions in regulated environments
  • Integrate token optimization into custom AI apps via SDK or proxy

📦 Installation

🧰 Requirements: Requires Python (pip), Node/npm, Rust, Homebrew, or Docker; no API key needed for local verification and simulation; optional API keys for proxy use.

Recommended: Python (pip)
pip install -U entroly
Node / npm
npm install -g entroly
Rust (source build)
cd entroly-core && cargo build --release --bin entroly-rs --features proxy
Homebrew
brew install juyterman1000/entroly/entroly
Docker
docker pull ghcr.io/juyterman1000/entroly:latest

Alternatively, use package runners:

Node / WASM runtime
npx -y entroly@latest --help
pnpm dlx entroly@latest --help
bunx entroly@latest --help
Complete Python runtime
uvx --from entroly entroly --help
pipx run --spec entroly entroly --help

🚀 Usage

Quickstart: auto-detect editor and wrap agent
pip install -U entroly && entroly go
Verify claims and simulate on your own repo (no API key)
cd /your/repo
entroly verify-claims
entroly simulate
Python SDK example
from entroly import compress, compress_messages, optimize
compressed = compress(api_response, budget=2000)
messages   = compress_messages(messages, budget=30000)
context    = optimize(fragments, budget=8000, query="fix the login bug")
CLI example
entroly compress response.json --out small.json
entroly recover sha256:0b957c79... --out restored.json

❓ FAQ

Do I need to change my code to use Entroly?

No. Entroly works with tools you already use like Claude Code, Cursor, Copilot, and 30+ others, and runs in the background.

Do I need an API key to try Entroly?

No. The commands entroly verify-claims and entroly simulate run locally with no API key and show real numbers on your own project.

How does Entroly ensure nothing is lost during compression?

Whatever Entroly sets aside is kept and can be pulled back exactly as it was, character for character, using content-addressed recovery.

Can I use Entroly with my own AI application?

Yes, you can use the SDK in Python (from entroly import compress, compress_messages, optimize) or set up a proxy by pointing your base URL to localhost:9377.

📊 Repository

Stars★ 443
Forks🍴 67
Open issues🐛 2
Last commit🕒 Sep 4, 2026
Created📅 Mar 2026
Language💻 Python
License⚖️ Apache-2.0

🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.