Every debugging session, architecture decision, and refactor you execute through AI coding tools produces valuable intellectual output. ReclaimLLM captures it automatically, indexes it with hybrid search, compresses token window waste, and lets you reuse context anywhere.
Capture Surface
Whether you write code in CLI agents, IDEs, terminal proxies, or browser tabs, ReclaimLLM normalizes every interaction into a single searchable timeline.
The richest capture surface. Native hooks record paired tool inputs and results, shell command outputs, and step-by-step file diffs as you code.
pip install rclm && rclm-hooks-installZero code changes required. Point your application or custom scripts to a local proxy address to log every request and response asynchronously.
export OPENAI_BASE_URL="http://localhost:8000/v1"Background extension captures web-based AI assistant conversations automatically, indexing browser chats alongside CLI engineering sessions.
Discovers and uploads past sessions that predate your hook installation so your history is complete from day one.
rclm-sync --allContext Efficiency
Frontier models charge heavily for long context windows. ReclaimLLM applies active compression rules locally before requests are sent to providers, saving up to 40% in tool-result token payloads on repetitive shell and read loops (16.5% overall in benchmark workloads).
Tracks file hashes and line ranges per session. Re-reads of unmodified files send compact line references instead of re-transmitting entire file bodies.
Filter hundreds of passing test lines across pytest, Jest, Vitest, Go, and Cargo. Only failing assertions and stack trace snippets are passed to the model.
Detects repeated failing tool calls to prevent runaway billing loops. Configurable brevity instructions cut conversational filler from model responses.
MCP Protocol
Install the RCLM MCP Server to query your past session memory directly from any MCP-enabled assistant (Claude Code, Cursor, Codex). Use tools like search_sessions, summarize_session, handoff, and transfer_session.
Terminal Visibility
The statusline installs automatically alongside native hooks. It displays real-time context window pressure, rate limits, peak-hour timing, active model, and git branch right in your terminal prompt.
ReclaimLLM uses hybrid semantic and exact keyword search across every captured session from Claude Code, Gemini CLI, Codex CLI, Cursor, Antigravity, local API proxy, and browser tabs. Developers can find debugging insights, regex solutions, and architectural decisions from weeks ago in seconds.
Yes. ReclaimLLM normalizes session records into a provider-neutral format, allowing you to carry forward complete debugging context and code histories between tools and models without re-explaining background context.
Running `pip install rclm` followed by `rclm-hooks-install` attaches native hooks to supported CLI coding agents. Hooks run locally before and after execution to capture paired tool calls, file diffs, and apply active context window compression.
The RCLM MCP Server exposes session memory tools (`search`, `summarize_session`, `handoff`, `transfer_session`) directly to MCP-compliant tools like Claude, Cursor, and Codex, allowing assistants to query your prior work directly.