#2020th of 27 in Memory
lossless-claw
OpenClaw plugin that replaces sliding-window compaction with a summary DAG over SQLite
- Stars
- 4.9k
- License
- MIT
- Last commit
- Nov 2026
- Last release
- Oct 2026
- Language
- TypeScript
4,236 on 21 Apr 20264,925 on 10 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Lossless-claw is a context-engine plugin for OpenClaw. It stores every message in SQLite, summarizes older chunks with a configured LLM, and condenses those summaries into a DAG. Each turn it assembles context from summaries plus recent raw messages, and exposes lcm_grep, lcm_describe and lcm_expand tools so agents can recover original detail. It also ships an lcm CLI, /lossless slash commands and a session migration tool.
Who it is for: OpenClaw users whose long-running agents lose context to compaction
Strengths
- Keeps every raw message in SQLite; summaries link back to source messages
- Agent tools (lcm_grep, lcm_describe, lcm_expand) recover details from compacted history
- Summary and expansion models configurable separately, with per-session spend guard
- Includes read-only lcm CLI, backup and doctor commands, and JSONL session backfill tool
Weaknesses
- Works only with OpenClaw 2026.9.2 or newer
- Summarization consumes LLM calls, adding provider cost and quota use
- Many tuning options (thresholds, fanout, deferred debt) make configuration involved
- Programmatic control API unavailable until the host supports the required endpoints
What it needs
- no GPU
- Docker
- Needs OpenClaw 2026.9.2+, Node.js 22+, SQLite, LLM provider configured in OpenClaw
- Models: Any provider configured in OpenClaw (e.g. openai/gpt-5.4-mini)
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Memory
See all 27| Rank | Project | Score |
|---|---|---|
| 1 | hindsightAgent memory server with retain, recall and reflect operations | 95 out of 100 |
| 2 | gbrainMemory store with sources and a knowledge graph for coding agents | 91 out of 100 |
| 3 | MemPalaceLocal verbatim memory store for AI agents with semantic search and MCP | 90 out of 100 |
| #4 | SupermemoryMemory and context engine for AI agents, with user profiles and hybrid search | 89 out of 100 |
| #5 | Mem0Long-term memory layer that stores and retrieves user facts for LLM agents | 88 out of 100 |