Press kit

Writing about LeanCTX?

Everything below is copy-ready and canonical: boilerplate in three lengths, a fact sheet sourced from the same single source of truth as this website, and the one correction we care about.

Open source. Open numbers. Open to questions.

Lean Context. Not Lean Cortex.

The one naming correction we care about.

LeanCTX is short for Lean Context: lean, precise context for AI agents instead of raw, noisy input. In prose, use LeanCTX; for the binary, package and command, use lean-ctx.

Early coverage occasionally misspelled the project as "Lean Cortex". If you spot it somewhere, we appreciate a correction. And if you are unsure about any claim, every number below is reproducible from the open-source repository.

Copy-ready, three lengths.

Use these verbatim or as a base. They are kept identical to the definitions on the homepage, /what-is-leanctx and in the repository README.

~50 words

LeanCTX (short for Lean Context) is the open-source context engineering layer that lets developers control what their AI can see. One local Rust binary decides what agents read, compresses what they send, remembers what they learn, guards what they touch, and proves what they save in a signed ledger. 60–90% fewer tokens, 30+ AI tools, zero telemetry.

~100 words

LeanCTX (short for Lean Context) is the open-source context engineering layer for AI agents. One local Rust binary decides what agents read (10 read modes, ~13-token cached re-reads) and compresses what they send (an optional local proxy, prompt-cache-safe), remembers what they learn across sessions, guards what they touch (PathJail, secret redaction, budgets) and proves what they save in an Ed25519-signed ledger. It works with 30+ AI coding tools (Cursor, Claude Code, OpenAI Codex, GitHub Copilot) via MCP and shell hooks, and embeds in any agent via a versioned /v1 API. Token savings of 60–90% are the receipt; context engineering is the product. Local use is free forever, enforced by CI.

~200 words

LeanCTX (short for Lean Context) is the open-source context engineering layer for AI agents. One local Rust binary decides what agents read (10 read modes, 60–90% fewer tokens, ~13-token cached re-reads), remembers what they learn (persistent sessions, knowledge graph), guards what they touch (PathJail, secret redaction, budgets, injection detection), proves what they save (Ed25519-signed ledger, reproducible benchmark) and replays what they saw (git-anchored, signed context snapshots you can restore or share); an optional local proxy compresses what they send — every request's system prompt, history and tool output, prompt-cache-safe on the wire. Compression — read-side and wire-side — is one of five subsystems, and every original stays locally retrievable. Works with 30+ AI coding tools via MCP and shell hooks; embeds in any agent via a versioned /v1 API with Python, TypeScript and Rust SDKs. Local use is free forever, enforced by CI. The project is written in Rust, parses 26 languages via tree-sitter AST, and ships 95+ shell compression patterns that shrink command output by 88–99%. Savings are recorded in an Ed25519-signed, hash-chained ledger on the user's machine and can be reproduced with the built-in benchmark (lean-ctx benchmark report). LeanCTX is local-first with zero telemetry: cache, memory and ledger are local files, and nothing leaves the machine unless the user configures it. The name expands to "Lean Context". The project is occasionally misspelled "Lean Cortex" in coverage, which the team actively corrects.

The numbers, from the source.

Sourced from the same numbers SSOT that renders this website. Every metric is reproducible with lean-ctx benchmark report on any repository.

NameLeanCTX (short for Lean Context); binary/CLI: lean-ctx
ClaimControl what your AI can see.
CategoryContext engineering layer for AI agents
LicenseApache-2.0 (open source; local use free forever, CI-enforced)
TechnologyOne Rust binary; tree-sitter AST for 26 languages
Token savings60–90% per read; ~13 tokens per cached re-read; 88–99% on shell output
Surface81-tool catalog (~13-tool lean default surface; full set on demand via ctx_call / ctx_expand), 10 read modes, 95+ shell patterns
Integrations30+ AI tools auto-detected (Cursor, Claude Code, Codex, Copilot, Windsurf, Cline…)
EmbeddingVersioned /v1 HTTP API (OpenAPI + /v1/capabilities), ctx_tools gateway, Python & TypeScript SDKs (GA, 14-check conformance kit), Rust SDK, WASM extension runtime, plugins
SecurityLocal-first, zero telemetry, PathJail, shell allowlist, secret redaction, OWASP-aligned injection screening
ProofEd25519-signed, hash-chained savings ledger; reproducible benchmark
GovernanceContext policy packs (baseline / strict-redaction / finance-eu / healthcare / open-source), first-class agent registry, org audit log, lean-ctx policy coverage
ComplianceEvidence Bundle v1 + offline leanctx-verify; EU AI Act / ISO 42001 / SOC 2 framework reports; Context Governance Benchmark (CGB)
FinOpsLeanCTX finops export (FOCUS 1.2 / CloudZero / Vantage), Datadog & Prometheus /metrics, net-of-injection (net_tokens_saved)
Linksleanctx.com · github.com/yvgude/lean-ctx

Logos, screenshots, and a human.

Everything needed to illustrate, verify, or follow up on a LeanCTX story.

Screenshots & demos

The homepage hero terminal and the live metrics page are press-usable as-is.

View live metrics

Deep background

What LeanCTX is, how the architecture works, and how every number is measured.

benchmark & methodology

Questions for a story?

We answer fast, provide reproducible numbers, and never ask to review your draft.

Support this project