What is Memory-DB CLAUDE.md?

Instructions for Memory-DB, a vector memory system for AI agents. It stores text in Qdrant, a database for similarity search, and uses llama.cpp to create the numerical representations used for searching.

Use it to store memories, search them by meaning, update or delete them, and inspect the four available MCP tools.

It defines how agents can save, find, change, and remove memories, including duplicate detection and search settings.

Memory-DB CLAUDE.md — Instructions file by cunzai97

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods

npx agentmods add instructions/cunzai97/memory-db/claude-md

Clone the repo

git clone --depth 1 https://github.com/cunzai97/Memory-DB

Per session 473 This file is loaded in full into every session.

When invoked 473 The same file — it is already loaded in full.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5	$0.00473	$0.00473
Opus 5	$0.00236	$0.00236
Sonnet 5	$0.00095	$0.00095
Haiku 4.5	$0.00047	$0.00047

Measured yesterday against content hash c01e39476799, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

Memory-DB CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 42 lines

What it actually says

Memory-DB — CLAUDE.md

Vector memory for AI agents: Qdrant (storage) + llama.cpp :8081 (embeddings). 4 MCP tools. Content must be <1024 tokens.

The Harness

Assertive pushback is non-negotiable. See global ~/.claude/CLAUDE.md § The Harness.

The 4 Tools (MCP)

`store_memory(content, tags?, dedup_threshold=0.85)`

Store text → vector. Dedup ≥ 0.85 replaces similar memories; set to 0 to disable. Returns {id, deduped}.

`get_memories(query, limit=5, min_score=0.5)`

Search by cosine similarity. Each hit increments recall_count. Returns sorted [{id, content, score, ...}] or [].

`update_memory(memory_id, content?, old_text?, new_text?, tags?)`

Update memory by ID. Two modes:

Full replace: content="..." — replaces entire content; re-encodes vector.
Partial replace: oldText="match" newText="replace" — finds exact substring and substitutes it; re-encodes vector.

At least one required. content and (oldText+newText) are mutually exclusive. Returns {updated: true, id, changes, update_type}.

`delete_memory(memory_id)`

Delete a memory by ID. Returns {deleted: true, id} or {deleted: false, id, error}.

Management CLI (not exposed to MCP)

memory-db-manage list [--limit N]           # list all memories
memory-db-manage export --path backups.json # JSON backup
memory-db-manage import --path backups.json # restore + re-encode
memory-db-manage rebuild                    # re-encode with current model
memory-db-manage purge ...                  # dry-run by default, add --execute to delete
memory-db-manage delete-all                 # destructive, requires confirm

Common Pitfalls

MCP failures are silent — always verify tool availability at session start.
Subagents can't use MCP tools — never delegate memory operations to background agents.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

yesterday First seen · 42 lines · 473 tokens per session scan A c01e39476799

Subscribe to this mod's changes

Memory-DB CLAUDE.md is an instructions file published in the GitHub repository cunzai97/Memory-DB (2 stars, last pushed 2mo ago), licensed MIT. It adds 473 tokens to every session, about $0.0024 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Data and AI Databases Memory and context

spec-kit AGENTS.md

AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.

github/spec-kit · 7,104 tokens

vscode buildNext.instructions.md

Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).

microsoft/vscode · 6,785 tokens

Memory-DB CLAUDE.md