Context Compression Python Packages

mcptoon

One zero-dependency CLI for every MCP server and agent skill. Token optimization, tool discovery and context compression: 71,929 -> 581 tokens (-99.2%, measured), schemas stay out of context. One config for Claude Code, Codex, Cursor, every agent. 341KB, pure Python. | 零依赖 CLI：管所有 MCP 工具与技能，工具发现省 99.2% token。

distil-llm

Compression with a quality contract — cache-aware, causally-pruned LLM context compression for agentic runtimes, certified non-inferior across 7 domains. Works with any SDK.

entroly-core

Cut AI context cost without trusting the compressor. Every reduction is reversible, byte-exact recoverable, and carries an auditable receipt. Local-first, works through proxy, MCP, SDK, or agent wrapper.

entroly

Cut AI context cost without trusting the compressor. Every reduction is reversible, byte-exact recoverable, and carries an auditable receipt. Local-first, works through proxy, MCP, SDK, or agent wrapper.

contextpress

Deterministic context compression for LLM chat, RAG, and agent pipelines

agent-coderag

CodeRAG: Lightweight semantic code search and distillation utility for AI coding agents. It solves the API knowledge gap via real-time local signature extraction and intent analysis. Optimized for token efficiency, it compresses codebase context into compact semantic summaries stored in a local DuckDB vector similarity index.

powermem

Not only memory but a full story.

specsmith

Lean governance for AI-assisted development: requirements tracking, test enforcement, epistemic context compression, auditable evidence, Grace local REPL, and coding-agent integrations.

furl-ctx

Context compression for AI agents — cut token usage costs with retrievable CCR compression. Rust core + Python API, Claude Code & Codex plugin, MCP server. PyPI: furl-ctx

powercontext

Not only memory but a full story.

sqz

Context compression for AI coding agents: compresses tool output before it enters the model, dedups repeats to 13-token refs. Claude Code, Cursor, Codex, Kiro, Zed, any MCP client. Rust, zero LLM calls.

graphsift

Token Saver for Claude, GPT-5 & Gemini. 80-150x code context reduction, F1 0.85. AST dependency graph, ranked context selection, 19 CLI compressors, MCP server, agent memory. Save LLM tokens — zero telemetry.

local-context-compiler

Cut LLM context and Claude Code sessions 40–80% without rewriting a byte. Offline, verbatim, with an audit receipt for every block dropped. CLI, MCP server and Claude Code/Open AI Plugin. MIT.

langstate

Inspectable context compression for LLM conversations: turn older history into visible scaffold state, keep recent turns verbatim, and check named facts with lexical receipts.

llm-zip

Self-hosted HTTP sidecar for LLM context compression. Reduce token costs 3–5× before calling any AI API — powered by LLMLingua-2 and MarkItDown. No proxy, no API keys, no GPU required.

fish-bridge-mcp

An economical and fuel(token) efficient AI tool, graph session memory for looong chat session

foldback-ai

Lossless context compression for LLM agents — fold it down, fold it back. Compresses tool outputs/logs before the model, prompt-cache-preserving, 39–82% fewer tokens. pip install foldback-ai

claw-compactor

14-stage Fusion Pipeline for LLM token compression — reversible compression, AST-aware code analysis, intelligent content routing. Zero LLM inference cost. MIT licensed.

justllm

Production LLM calls. Just the three lines. Cross-provider fallback, native caching, and reversible context compression on by default.

foveance

A lossless token-optimization codec for LLM agents — cut input tokens up to 82%, nothing dropped, every fact kept. Drop-in proxy + one-line Python API for Claude Code, Codex, Cursor, OpenAI, Anthropic & Ollama. pip install foveance

headroom-winnow

Winnow: separate the grain from the chaff in your coding agent's tool output. A task-aware pruner for Headroom, powered by Squeez models.

ctxprune

Context compression for AI agents: LLMLingua-2 successor that keeps IDs, paths and numbers intact. 68.8% vs 53.4% QA at 50% kept.

powermem-mcp

Not only memory but a full story.

mnemosyne-mcp

State aware knowledge compression, ingestion, and hybrid retrieval engine. Zero dependencies. Sub-100ms queries.