Wasted tokens
Expensive models handle tasks a more targeted path could solve. Without compression, your agent sends 10× more context than it needs.
Your agents re-read everything. Every single time.
90% of every prompt is wasted context. LeanCTX compresses file reads, shell output, and search results — so your agent gets only what matters.
Starred by engineers at
GoogleMicrosoftAmazonIBMRed HatTencentAutodeskDynatraceDellVTEXZühlke
The Cold-Start Tax. Paid every session.
Every time an agent starts, it re-reads the same files, re-runs the same commands, and re-discovers the same architecture. The pattern repeats because context and memory never persist.
Blind token spend
10,000 tokens go in — but which lines mattered? Without visibility, you cannot optimise what you cannot measure.
Knowledge lost per session
Decisions, patterns, architecture insights — gone when the session ends. The next agent starts from zero.
Context in. Intelligence out.
What LeanCTX engineers.
Four capabilities that turn raw file access into precise, persistent, verifiable context delivery.
01
Read Compression
10 read modes powered by tree-sitter AST analysis across 27 languages. Your agent sees structure, not noise.
Learn more02
Shell Intelligence
95+ compression patterns for git, cargo, npm, docker, and more. Thousands of lines become the signal that matters.
Learn more03
Persistent Memory
Knowledge graph retains decisions, patterns, and architecture insights across sessions. What was learned once is never lost.
Learn more04
Agent Coordination
Agent bus, shared sessions, and handoff protocols. Multiple agents work together without duplicating context.
Learn moreLess in. More out.
Before your agent reads a file, LeanCTX selects the relevant context: AST-aware extraction, semantic deduplication, intelligent truncation. Fewer tokens, same understanding.
Get StartedEvery token accounted for.
The signed savings ledger tracks compression ratios, token counts, and cost savings — cryptographically verified. Not a claim, a proof.
Get StartedLearned once, available forever.
The knowledge graph connects code patterns, architectural decisions, and session insights. New agents inherit what previous sessions discovered.
Get StartedBuilt in the open. Adopted globally.
Common questions.
What developers want to know before adding LeanCTX to their workflow.
LeanCTX uses tree-sitter AST analysis to understand code structure. It preserves signatures, types, and logic while removing whitespace, comments, and boilerplate. The 10 read modes let you choose the right level of detail for each task.
Cursor, Claude Code, Codex CLI, Windsurf, Cline, and GitHub Copilot. LeanCTX runs as an MCP server that any compatible agent can connect to.
No. LeanCTX runs entirely on your machine. Files are read, compressed, and delivered locally. Nothing leaves your filesystem.
LeanCTX is open source and free for individual developers. Team features and cloud sync are available through Thinkery.
One command.That's it.
LeanCTX installs in 30 seconds and works with your existing agent. No configuration changes, no workflow disruption. Just better context.
curl -fsSL https://leanctx.com/install.sh | sh