By 42U
Permanent, graph-based memory for Claude Code that persists knowledge across sessions using SurrealDB and BGE-M3 embeddings. Automatically extracts, validates, and retrieves project insights, documents, and decisions, with tools for backup, export, health monitoring, and local graph visualization.
You are a LaqrumCode memory processing agent. Your job is to process pending knowledge extraction work from previous sessions, turning raw conversation data into structured knowledge.
You are a LaqrumCode memory processing agent. Your job is to process pending knowledge extraction work from previous sessions, turning raw conversation data into structured knowledge.
You are an introspection specialist for LaqrumCode. Your job is to explore and present the agent's self-knowledge and identity.
Activate periodically (not per-session) to sample concepts from the graph and verify their claims against current state — code paths, recent sources, user-provided ground truth. Flag stale concepts for supersede-stale. Without this, knowledge rot poisons retrieval silently.
Activate when an insight worth keeping emerges mid-session — a surprising finding, a reusable pattern, a user correction, or a decision with rationale. Use this to write knowledge into the laqrumcode graph immediately rather than waiting for end-of-session daemon extraction (which is batch-only and can have hours of lag).
Activate when the user wants to distill lasting knowledge from ANY source — PDF, web article, pasted text, codebase file, transcript, markdown doc, or book chapter. Source-agnostic replacement for extract-pdf-gems. Same workflow core; source-type-aware quality rules.
Activate when the user asks to extract, mine, distill, or "pull gems" from a PDF, academic paper, book chapter, or research document into the laqrumcode memory graph. Also activate for phrases like "read this paper and remember it", "dense knowledge gems", "pull the insights", or when the user points at a PDF and asks to study it.
Activate when the user asks about state, history, prior work, codebase knowledge, or anything where injected memory should inform the response. Triggers include "what do you know about", "remember", "earlier", "last time", "we discussed", "in this codebase", "prior work", "previously", or any direct question about project/session state.
Matches all tools
Hooks run on every tool call, not just specific ones
Admin access level
Server config contains admin-level keywords
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
Has parse errors
Some configuration could not be fully parsed
Has parse errors
Some configuration could not be fully parsed

Graph-backed permanent memory for Claude Code.
Quick start | Architecture | Configuration | Troubleshooting | Development
LaqrumCode gives Claude Code a persistent, queryable memory that grows with every session. It extracts concepts, causal chains, corrections, preferences, decisions, and skills from conversations automatically — then retrieves them with a multi-stage pipeline (BGE-M3 vectors → learned ACAN scoring → cross-encoder reranking, 98.2% R@5) backed by a SurrealDB graph running locally. After enough sessions, the agent earns an emergent identity (soul graduation) grounded in its actual working history.
| Capability | Stock Claude Code | With LaqrumCode |
|---|---|---|
| Memory | File-based, per-project, manual | Graph DB, cross-session, automatic |
| Context window | Sliding, lost on /clear or session end | Retrieval-augmented from prior turns and concepts |
| Knowledge extraction | None | Concepts, causal chains, monologues, corrections, preferences, artifacts, decisions, skills, reflections |
| Procedural memory | None | Skills mined from successful workflows, surfaced when preconditions match |
| Retrieval quality | None (no memory to retrieve) | Vector search → ACAN reranker → cross-encoder rerank → graph expansion (98.2% R@5 on LongMemEval) |
| Identity | Stateless on every turn | Earned soul after a graduation gate (volume + quality thresholds) |
Install the plugin, set up the system prompt, and open a session. The daemon provisions SurrealDB, the embedding model, and everything else on first run.
| Tool | When required |
|---|---|
| git | Always (Claude Code uses it to clone the marketplace repo) |
| Node.js >= 18 + npm | Only for the JS fallback path. Most users on linux-x64/arm64, macOS x64/arm64, win-x64 get the SEA binary and don't need Node. |
Quick installs (only if you need Node + git for the fallback path):
brew install node gitwinget install OpenJS.NodeJS.LTS Git.Git then restart your terminal AND Claude Code so the new PATH is picked up.apt install nodejs npm git) or nvm.CLI:
/plugin marketplace add 42U/laqrumcode
/plugin install laqrumcode@laqrumcode-marketplace
The plugin system is shared across CLI and IDE extensions. Open a Claude Code session in your IDE, then:
Type /plugins and press Enter (shows Manage Plugins)
Go to the Marketplaces tab
Type 42u/laqrumcode in the input field and click Add

Switch to the Plugins tab
Toggle laqrumcode@laqrumcode-marketplace on

LaqrumCode ships two template files that teach Claude how to use the memory graph. Set both up before your first session.
# Find the plugin cache (installed by step 1)
LAQRUMCODE_DIR=$(ls -d ~/.claude/plugins/cache/laqrumcode-marketplace/laqrumcode/* 2>/dev/null | head -1)
# Copy the system prompt (appended every session via CLI flag)
cp "$LAQRUMCODE_DIR/templates/laqrumcode.txt" ~/.laqrumcode-prompt.txt
npx claudepluginhub 42u/laqrumcode --plugin laqrumcodeOpenLTM — Long-Term Memory for AI coding agents: semantic search, context injection, session learning
Persistent semantic memory for Claude Code — auto-injects relevant context at session start, captures git commits, test results, and research via hooks, and provides 30+ MCP tools for memory management.
Multi-tiered memory and knowledge base with semantic search, auto-compaction, and built-in evaluation. Works across Claude Code, Copilot CLI, OpenCode, Cline, and Cursor.
Persistent memory across Claude Code sessions using Cognis
Persistent memory for Claude Code. Capture work across sessions and recall relevant context.
Persistent memory system for Claude Code with SQLite, FTS5, and vector search