juyterman1000/entroly

Accuracy Retention

Does Entroly compression degrade LLM answer quality? No. All 6 confidence intervals overlap baseline.

Model: gpt-4o-mini · Budget: 50K tokens · Wilson 95% CI · Reproduce: python -m bench.accuracy --benchmark all

Average retention 101.7% — accuracy is statistically indistinguishable from raw context across all benchmarks (zero degradation). For live codebase token compression (85–94% reduction), see Context Selection Quality below.

Context Selection Quality

19-fragment corpus · 300-token budget · 3 real-world queries · Reproduce: entroly benchmark

Entroly sees 8.7 modules where TOP-K sees 3.7 — it includes auth, payments, AND rate limiting. TOP-K misses the rate limiter. Full methodology, CIs, and reproduce commands →

Research

Entroly implements six research-grade algorithms with production implementations:

Read the full research documentation · Cite Entroly

Integration hub

Use Entroly at the SDK, framework, proxy, MCP, plugin or agent boundary. A listed name is not automatically a claim that hosted subscription inference is intercepted; provider-bound savings exist only when the request traverses an Entroly-controlled route.

Open the complete verified integration and operations hub →

What is Entroly? (in plain English)

AI coding assistants have a memory limit. Hand one your whole codebase and it gets slow, expensive, and distracted — like giving someone a 500-page manual when they only needed page 47.

Entroly finds page 47.

It sits between your code and the AI, reads everything, and passes along only the parts that matter for the question actually being asked. Three things make that safe to do:

Do I need to pay for anything to try it? No. The two commands in the Install section below run on your own machine, with no API key, and show you real numbers on your own project before you connect anything paid. (They will install the native engine from PyPI if it is missing — see the note under Install.)

Install

Not sure which one? Pick Python. It's the complete version and what most people use. The others are alternate ways to run the same engine. | Platform | Install | What you get | |---|---|---| | 🐍 Python (pip) — recommended | pip install -U entroly | Everything: the command-line tool, the server your AI editor talks to, and the code library | | 📦 Node / npm | npm install -g entroly | The same engine, nothing Python required | | 🦀 Rust (source build) | cd entroly-core && cargo build --release --bin entroly-rs --features proxy | One self-contained program, no Python or Node needed | | 🍺 Homebrew | brew install juyterman1000/entroly/entroly | The command-line tool on macOS/Linux | | 🐳 Docker | docker pull ghcr.io/juyterman1000/entroly:latest | Runs in a container, nothing installed on your machine |

Prefer a package runner instead of a global install? These commands use the same published artifacts in an isolated tool cache:

# 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

The Node commands provide the local WASM CLI. The Python commands provide the complete CLI, SDK, MCP, proxy, verification, and native-engine path described above. Entroly's release workflow smoke-tests all five runners against the exact version before a release is considered complete.

Now check that it worked — free, no API key:

cd /your/repo
entroly verify-claims
entroly simulate

Both run locally. Neither one calls an AI or costs anything.

One exception to "offline": if the native engine is missing, Entroly installs it from PyPI before measuring, because without it selection cannot read your query and any savings figure would be budget arithmetic rather than a measured result. That is the only outbound call these commands make, it is a package install and nothing about your code leaves the machine, and it does not happen when the engine is already present. Set ENTROLY_NO_SELF_HEAL=1 to disable it — Entroly then reports the figure explicitly labelled as unearned.

Extras (entroly[proxy], entroly[native], entroly[full]), the standalone Rust binary, and uninstall steps: Engine & install options.

Contributing from source? Follow the reproducible development setup. Local installation and the normal test suite need no API key; .env.example documents only optional workspace, offline, provider, and proxy settings.

Quickstart — by how you work

Just want it working? pip install -U entroly && entroly go — that's the whole thing. It finds your editor, sets itself up, and shows you a before/after dashboard. The rest of this table is for specific setups. | Your situation | Do this | What it gets you | |---|---|---| | 🟢 "I just want it on." (pip / Python user) | pip install -U entroly && entroly go | Auto-detects your editor, wraps your agent, opens a dashboard showing tokens before and after | | "I use Node, not Python." (npm user) | npm install -g entroly && entroly init | Same engine, nothing Python required | | "I want one binary, no runtime." (Rust user) | cargo build --release --bin entroly-rs --features proxy (from entroly-core/) | A single native program with no dependencies | | "I use Claude Code, Codex, Gemini CLI, or VS Code agent plugins." (plugin user) | Install the Entroly plugin/extension for that host, submit one prompt, then run entroly activation status --json | A trusted prompt hook performs bounded local selection before planning; a receipt proves the hook ran | | "I use Cursor with third-party configs enabled." | entroly activation install --host cursor --project . | Merges a reversible Claude-compatible prompt hook; native Cursor MCP remains advisory | | "I use Kiro IDE 1.x or CLI 3.x." | entroly activation install --host kiro --project . | Installs a reversible project PromptSubmit hook whose stdout is added to agent context | | "I use another MCP host." | entroly attach create --client claude --project . --ttl 4h --install or the client-specific command in the compatibility matrix | Scoped Entroly tools and receipts; the model can still skip MCP unless the host has a verified lifecycle hook | | "I'm building my own app in Python." (SDK user) | from entroly import compress, compress_messages, optimize | Call it straight from your code, anywhere you assemble a prompt |

Cursor MCP users can also use this one-click install link (no marketplace account required): Add Entroly to Cursor. | "I have an API key and my own app." (proxy user) | entroly proxy → point ANTHROPIC_BASE_URL / OPENAI_BASE_URL / GOOGLE_GEMINI_BASE_URL at localhost:9377 | Every request gets optimized on the way past — no code changes on your side |

Runaway-session rescue — automatic on the proxy, callable everywhere else. When a long agent session approaches the provider's context limit, bulky tool output is compacted in flight: no manual /compact, the prompt prefix stays byte-stable so your warm provider cache survives, and every omitted span is recoverable. The proxy does it for you because it sees the outbound request. Anywhere else — pip, SDK, a provider-SDK wrapper, or an MCP host that passes its transcript — hand the conversation over and get the same policy: from entroly import rescue_session. entroly capabilities reports which protections apply to how you are running. See session rescue.

Why bother: less unnecessary context reaches the model (lower bill, less distraction for the model), nothing is silently lost (every drop is recoverable and receipted), and you can prove it — entroly verify-claims and entroly simulate show real numbers on your own repo before you connect a paid key.

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")

entroly compress response.json --out small.json
entroly recover sha256:0b957c79... --out restored.json

Full setup paths for every agent, IDE, and CI use case: Get started in depth · Command reference.

See it work in 30 seconds

Not mocked recordings — each video is rendered from a checked-in command that verifies its source artifact before printing a number.

Full protocols, sample sizes, and every caveat: docs/BENCHMARKS.md.

Benchmarks

The question that matters: if you send less, does the AI start getting things wrong? These are standard public tests, run with and without Entroly.

How to read this: Retention is how well the AI still answered — 100% means it did just as well on far less text. Token savings is how much less was sent (and therefore paid for). Measured with gpt-4o-mini; intervals are Wilson 95% CIs.

Being straight with you: look at the SQuAD 2.0 row — accuracy went down (80% → 72%). Compression is a trade, not magic, and it doesn't win everywhere. That's why entroly simulate exists: run it on your own project and see your own numbers before you commit to anything.

Hallucination detection (WITNESS, local, no API): 84.92% accuracy / 0.7976 AUROC on 20,000 HaluEval-QA decisions — within the reported uncertainty of gpt-4o-mini as an API judge on the same shared sample.

Frozen evidence-selection benchmark (opt-in PRISM-R research prototype, not the default compressor): a disagreement guard kept the answer-bearing passage in 298 of 300 cases while selecting an average of 1.02 of 16 passages (paired exact McNemar p=0.21875 vs. BM25 alone) — this experiment measures retrieval of the known-answer passage, not generated-answer quality. Full protocol: PRISM-R neural evidence frontier.

Recovery, latency, and head-to-head frontier results are in docs/BENCHMARKS.md with raw artifacts linked. None of these numbers are a universal or production-savings guarantee for your workload — reproduce them on your own repo with entroly simulate and entroly value.

Features

• Picks first, shrinks second — it works out which files actually answer your question, then compresses them.
• Gives you the original back, exactly — anything left out can be restored character-for-character and checked against a fingerprint.
• Shows its work — a receipt for every decision: what was kept, what was left out and why, and what risk remains.
• Fact-checks answers — compares what the AI said against the evidence it was given, on your machine, without paying for a second AI call.
• Doesn't wreck your caching — keeps the unchanging parts of your prompt stable so your provider's discount for repeated text still applies.
• Rescues sessions before they crash — when a conversation grows too big, it trims recoverable output instead of letting the provider reject the request mid-task.
• Can route cheap work to cheap models — optional and fail-closed when uncertain.
• Cross-agent shared memory — Claude, Codex, Cursor, and Gemini can read and write the same compressed context store with automatic SimHash deduplication and agent provenance tracking.
• Output token reduction — effort-based routing classifies query complexity and steers model verbosity, reducing output tokens alongside input tokens.
• Shell hook compression — transparent CLI output compression for git, npm, cargo, docker, pytest, kubectl, and terraform. Preserves errors and warnings, strips progress bars and boilerplate.
• Image compression — 40-90% reduction on screenshots and diagrams for vision API calls, with optional OCR text extraction.
• Failure mining — entroly learn --deep mines session data for recurring failure patterns and writes corrections to CLAUDE.md, .cursorrules, and other agent configs.

Runs as a CLI, Python/TypeScript SDK, MCP server, HTTP proxy, or library import. Full surface map: docs/product-surface.md. Architecture and Rust internals: docs/DETAILS.md.

How Entroly compares

Most context tools compress and hope. Entroly is an auditable context control plane — every selection is receipted, every compression is reversible, and every claim is verifiable.

Compression-quality frontier (September 2026)

Every tool measured on its own published benchmarks. Different datasets — not apple-to-apple — but the compression-retention tradeoff is comparable.

Sources: PointFive 2026 guide, SuperCompress benchmarks, published tool docs. "Baseline D" is anonymized per project policy. Entroly numbers link to frozen JSON artifacts in docs/BENCHMARKS.md.

What only Entroly has

What's different: Entroly is the only tool that combines optimal selection (knapsack solver), auditable receipts (byte-offset fragments, SHA-256 digests, inspectable omissions), verification (WITNESS grounding, EICV hallucination detection), and zero-cost Bayesian learning (5D PRISM weights). Each competitor has one piece of this; Entroly has the full stack.

Works with your stack

Install the public Codex plugin from the Entroly repository:

codex plugin marketplace add juyterman1000/entroly --ref main
codex plugin add entroly@entroly-public

Restart Codex, review and trust the hook, then run entroly activation status --json after a task. The marketplace installs the local Node/WASM runtime with the plugin; the model does not have to remember to call an MCP tool before Entroly runs. A receipt proves that the hook executed and selected local context or made an explicit no-match decision. It does not prove token or cost savings without a matched provider-bound baseline.

Install the same public repository as a Gemini CLI extension:

gemini extensions install https://github.com/juyterman1000/entroly --ref main --consent

Restart Gemini CLI after installation. The repository root contains gemini-extension.json and GEMINI.md, so the command works without navigating into an integration subdirectory.

For VS Code or Kiro, download the entroly-vscode-*.vsix asset from the latest GitHub release, then install it with Extensions: Install from VSIX or code --install-extension. The extension is self-contained and does not require an API key.

JetBrains AI Assistant users can add the same server globally at Settings → Tools → AI Assistant → Model Context Protocol (MCP):

{
  "mcpServers": {
    "entroly": {
      "command": "npx",
      "args": ["-y", "entroly-mcp@1.0.84", "serve"],
      "env": {
        "ENTROLY_NO_DOCKER": "1",
        "ENTROLY_MCP_PASSIVE": "1",
        "ENTROLY_MCP_PROFILE": "public",
        "ENTROLY_MAX_FILES": "200"
      }
    }
  }
}

The repository also ships a free, open-source JetBrains plugin that guides this setup from Tools → Configure Entroly for AI Assistant, checks the local runtime on request, and keeps the evidence boundary visible. See extensions/jetbrains.

MCP marketplace and plugin manifests select the compact public profile so agents see the core context, receipt, continuity, recovery, and verification tools first. A direct entroly serve invocation remains backwards compatible and exposes the full tool surface. You can choose either behavior explicitly with ENTROLY_MCP_PROFILE=public or ENTROLY_MCP_PROFILE=full.

The MCP path is provider-neutral: the host can use OpenAI, Anthropic, Google, Mistral, DeepSeek, Kimi, GLM, or a local model. There is no separate plugin marketplace for each model provider; the host's MCP or extension contract is the integration boundary.

Hook enforcement belongs to the host, so it is independent of whether that host runs an OpenAI, Anthropic, Gemini, Kimi, DeepSeek, Mistral, or GLM model. Status describes integration depth, not a savings guarantee. Provider-observed savings require requests to traverse an Entroly proxy route. Entroly does not claim interception of GitHub-hosted subscription inference on Copilot's native path. Full compatibility matrix: docs/agent-compatibility.md.

Entroly carries verified metadata for current models from OpenAI, Anthropic, Google, Meta, and others. It auto-discovers local Ollama models. Model-specific details: docs/DETAILS.md.

When to use it · when to skip it

Great fit: large repos where the agent only sees a few files at a time · chatty multi-turn agents · anywhere you want answers checked against evidence · cutting a real, growing AI bill.

Skip it: tiny repos or short prompts that already fit the budget · judgment-heavy tasks where you always want the full flagship model.

More commands

For evidence-led optimization rather than a synthetic savings estimate:

entroly learn --history --json
entroly shrink -- pytest -q
entroly trial --experiment checkout-fix --arm baseline -- codex exec "fix the checkout test"
entroly trial --experiment checkout-fix --arm optimized -- codex exec "fix the checkout test"
entroly trial --report checkout-fix
entroly browser https://example.com --query "billing settings"
entroly response set evidence --scope project

Trials run one explicitly selected arm at a time so a stateful or paid agent task is never repeated implicitly. Response contracts shape agent instructions; they do not truncate responses or count as measured savings. Browser and command reductions keep exact local recovery handles and pass through when their safety gates cannot be met.

For teams that need to say who an agent is and what it was allowed to do:

entroly govern status                          # identity, policies, audit chain
entroly govern policy check write --risk high  # evaluate one authorization
entroly govern audit verify                    # exit non-zero on a broken chain

Authorization is deny-by-default and every denial names the policy and the reason it gave. audit verify checks that recorded entries were not altered after the fact — it does not prove every action was recorded, and govern status reports the state of the local control plane only, not an attestation that each agent action passed through it. Identity tokens are unsigned unless ENTROLY_IDENTITY_KEY is set, and the credential is never printed.

Also available: entroly wrap, entroly unwrap, entroly serve, entroly daemon, entroly dashboard, entroly demo, entroly capabilities, entroly ingest, entroly select, entroly receipt, entroly explain, entroly context-commit, entroly proof, entroly benchmark, entroly cache, entroly ravs, entroly perf, entroly batch, entroly usage. Full description: command reference.

Common questions

Will this change my code or my files?

Does my code get uploaded anywhere?

What if it leaves out something important?

How much money will this actually save me?

I'm not a developer. Can I use this?

Something broke / I'm stuck.

Cross-agent shared memory

Content-addressed store with SimHash deduplication and BM25 search. Multiple agents (Claude Code, Codex, Cursor) write and query the same knowledge base with provenance tracking.

from entroly import shared_memory_write, shared_memory_search
shared_memory_write("Auth uses JWT with RS256", agent_id="claude-code", tags=["auth"])
results = shared_memory_search("authentication tokens")  # finds it, from any agent

Output token reduction

Three-layer pipeline: effort classification steers verbosity directives, max_tokens budgets cap generation, and post-generation distillation trims filler. A "yes/no" query gets 150 max tokens; a detailed architecture review gets 16,384.

Shell hook compression

Command-specific patterns for git, npm, cargo, docker, pytest, kubectl, and terraform strip progress bars, deprecation warnings, and boilerplate while preserving errors and key results. Full output is recoverable via content-addressed handles.

entroly hook install     # adds transparent compression to your shell
entroly hook status      # shows which shells have the hook

Failure mining

entroly learn --deep mines PRISM feedback, vault beliefs, evolution daemon, and checkpoint data for recurring failure patterns, then generates corrections for agent config files.

How Entroly compares

Entroly is the only tool that combines optimal selection (knapsack solver with provable guarantees) with auditable receipts (byte-offset fragments, SHA-256 digests, inspectable omissions) and verification (WITNESS grounding checks, EICV hallucination detection).

Documentation

• Architecture & internals — Rust modules, compression pipeline, provenance, command reference.
• Benchmarks — every number, protocol, artifact, and caveat.
• Agent compatibility — every supported client and its authentication boundary.
• Limitations — where Entroly helps, where it passes through, what it doesn't guarantee.
• Product surface map — CLI, SDK, MCP, proxy, verification, memory, security.
• First-run trust — what to run before wiring a paid model key.
• Cookbook — copy-paste recipes.
• Discussions · Issues