Metadata-Version: 2.4
Name: entroly
Version: 1.0.61
Summary: Local context OS for AI coding agents: token optimization, MCP tools, receipts, exact recovery, verification, and SDK helpers.
Author-email: entroly <fastrunner10090@gmail.com>
License-Expression: Apache-2.0
License-File: LICENSE
License-File: NOTICE
Keywords: agentic-ai,claude-code,context-compression,context-control-plane,context-optimization,context-receipts,copilot,cursor,entropy,github-copilot,hallucination-detection,knapsack,llm,llm-proxy,local-context-os,mcp,mcp-server,reduce-llm-costs,token-reduction,token-saving
Classifier: Development Status :: 4 - Beta
Classifier: Intended Audience :: Developers
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Programming Language :: Python :: 3.14
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Classifier: Topic :: Software Development :: Libraries :: Python Modules
Requires-Python: >=3.10
Requires-Dist: mcp<2,>=1.6.0
Provides-Extra: benchmark
Requires-Dist: tiktoken<0.10,>=0.9; extra == 'benchmark'
Provides-Extra: full
Requires-Dist: cryptography>=42; extra == 'full'
Requires-Dist: entroly-core<2,>=1.0.61; extra == 'full'
Requires-Dist: httpx>=0.27; extra == 'full'
Requires-Dist: starlette>=0.37; extra == 'full'
Requires-Dist: uvicorn>=0.30; extra == 'full'
Provides-Extra: native
Requires-Dist: entroly-core<2,>=1.0.61; extra == 'native'
Provides-Extra: neural
Requires-Dist: sentence-transformers<6,>=5.1; extra == 'neural'
Provides-Extra: neural-benchmark
Requires-Dist: datasets<5,>=4; extra == 'neural-benchmark'
Requires-Dist: sentence-transformers<6,>=5.1; extra == 'neural-benchmark'
Provides-Extra: proxy
Requires-Dist: httpx>=0.27; extra == 'proxy'
Requires-Dist: starlette>=0.37; extra == 'proxy'
Requires-Dist: uvicorn>=0.30; extra == 'proxy'
Provides-Extra: receipt-proof
Requires-Dist: cryptography>=42; extra == 'receipt-proof'
Provides-Extra: test
Requires-Dist: pytest<9,>=8; extra == 'test'
Description-Content-Type: text/markdown

<p align="center">
  <img src="https://raw.githubusercontent.com/juyterman1000/entroly/main/docs/assets/entroly_wordmark.svg" width="760" alt="Entroly">
</p>

<p align="center"><b>Know exactly what your AI agent saw.</b></p>

Entroly is a local context control plane for AI coding agents. It selects the
highest-value evidence under a token budget, records what was selected and
omitted, keeps compressed context recoverable, and produces verifiable Context
Commits and receipts.

## Install

```bash
pip install -U entroly
```

Run the local, no-key verification path:

```bash
entroly verify-claims
entroly simulate
```

## MCP server

For an MCP client, register the installed `entroly` command with no arguments.
When an MCP client launches it with a stdio pipe, Entroly starts the installed
Python server directly:

```bash
entroly
```

Or register a package runner, also with no `serve` argument:

```bash
uvx --from entroly entroly
npx -y entroly-mcp
```

`entroly serve` is a different deployment path: it uses the Entroly Docker
image by default. For the installed Python runtime in an interactive shell, use
`ENTROLY_NO_DOCKER=1 entroly serve` on macOS/Linux or set
`ENTROLY_NO_DOCKER=1` in the client environment.

Entroly works with GitHub Copilot in VS Code, Claude Code, Cursor, Windsurf,
Cline, Continue, Zed, and other MCP-compatible clients.

### GitHub Copilot / VS Code

Create `.vscode/mcp.json`:

```json
{
  "servers": {
    "entroly": {
      "type": "stdio",
      "command": "uvx",
      "args": ["--from", "entroly", "entroly"]
    }
  }
}
```

External MCP galleries can lag a release. Direct stdio registration above is
the canonical setup; confirm a gallery entry's package version and validation
status before relying on it.

### Claude Code

```bash
claude mcp add entroly -- uvx --from entroly entroly
```

### Generic MCP configuration

```json
{
  "mcpServers": {
    "entroly": {
      "command": "uvx",
      "args": ["--from", "entroly", "entroly"]
    }
  }
}
```

## What Entroly adds

- **Context selection** under explicit token and cost budgets
- **Context Commits** linking selected, omitted, and recoverable evidence
- **Context Receipts** for replay, audit, and omission explanations
- **Exact recovery** of compressed fragments through stable handles
- **Local verification** through WITNESS and receipt checks
- **Context Check** coverage evidence for changed files and CI risk gates
- **Pure-Python base runtime**, optional Rust acceleration, and a separate npm/WASM runtime
- **Local-first operation** with no outbound analytics by default

## Links

- Repository: https://github.com/juyterman1000/entroly
- Documentation: https://juyterman1000.github.io/entroly/
- PyPI: https://pypi.org/project/entroly/
- npm runtime: https://www.npmjs.com/package/entroly
- npm MCP bridge: https://www.npmjs.com/package/entroly-mcp
- Public evidence policy: https://github.com/juyterman1000/entroly/blob/main/docs/public-evidence.md

## MCP Registry identity

`mcp-name: io.github.juyterman1000/entroly`

Apache-2.0 licensed.
