Lossless Claw is a plugin for OpenClaw that replaces the default context management strategy with a lossless summarization system.
The problem it solves is that conversations grow beyond a language model's context window, forcing agents to discard older messages. Lossless Claw preserves every message in a SQLite database while keeping active context within token limits. It works by summarizing chunks of older messages into nodes that form a directed acyclic graph, then assembling context each turn by combining summaries with recent raw messages. The tool provides agents with search and recall capabilities through tools like lcm_grep, lcm_describe, and lcm_expand so they can drill into compacted history and recover original details when needed.
Adoption suits projects using OpenClaw where conversation history matters and you want agents to reference past interactions without losing information. The tool is particularly valuable for long-running conversations or applications where forgetting context is undesirable. It installs an agent-oriented shell CLI for managing conversations and includes a bundled skill for in-session operations. The shell CLI reads the conversation database without modifying it, with safety constraints on configuration writes. Output defaults to JSON format with bounded keyset pagination for list commands.
The project shows consistent maintenance with regular commits addressing both feature development and bug fixes. Pull requests receive timely review and merge activity. The codebase demonstrates attention to code quality through testing infrastructure and documentation updates. Issue tracking reflects active engagement with user-reported problems and feature requests.
