LeanCTX

LeanCTX helps developers control the information their AI coding tools receive. It runs locally between those tools and project files or command output, while your existing application keeps control of the model and workflow.

• Focused context: Search, code outlines and compressed file reads or command output reduce repeated material, with references to recover details when retained sources and permissions allow.
• Access and content controls: Restrict file access, limit context size, and warn, redact or block matching content on supported integration paths.
• Memory across sessions: Keep local task records, findings and decisions, and use code and knowledge graphs to find related information across chats.
• Usage and audit records: Inspect context activity in a browser dashboard and export recorded source references, policy decisions and usage evidence.

The open-source Engine uses the Apache 2.0 license and runs on macOS, Linux and Windows. It connects through CLI, MCP, shell hooks and supported local proxy paths, including integrations with Claude Code, Cursor and Codex. The SDK has separate licensing terms.

Task records and context processing stay local; model requests go to the provider chosen by your AI application. Controls depend on the integration and settings, and file access restrictions aren't an operating-system sandbox. Context-only records can't verify a later model call. Claude request compression through the proxy requires an Anthropic API key; subscription OAuth doesn't support that proxy path.