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.
npx agentmods add instructions/quequiere/graph-mem/claude-mdgit clone --depth 1 https://github.com/quequiere/graph-memWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/quequiere/graph-mem/claude-md)<a href="https://agentmods.dev/instructions/quequiere/graph-mem/claude-md"><img src="https://agentmods.dev/badge/instructions/quequiere/graph-mem/claude-md.svg" alt="Measured on agentmods" height="20"></a>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 | $0.02842 | $0.02842 |
| Opus 5 | $0.01421 | $0.01421 |
| Sonnet 5 | $0.00568 | $0.00568 |
| Haiku 4.5 | $0.00284 | $0.00284 |
Grade B, and why
graph-mem CLAUDE.md scanned grade B with 1 finding 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 3d ago.
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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
Configured in `~/.claude/settings.json`: How it starts
The opening of the file, as written. The whole thing — 228 lines — stays where its author put it; the contents beside it link to each section on GitHub.
graph-mem - Project Tracking
Project Overview
graph-mem is a Claude Code plugin that provides persistent, intelligent memory using Graphiti's knowledge graph. Unlike flat memory stores, it leverages entity relationships, temporal awareness, and semantic search to build a rich understanding of the developer over time.
Current Status
Working end-to-end on branch feat/openrouter-graphiti-patch. Stack: Neo4j 5.26 + patched Graphiti + OpenRouter (gemma-4-26b + qwen3-embedding-8b).
Design Spec
Full design document: .docs/specs/2026-04-05-graph-mem-design.md
Architecture
Claude Code (stdio) → graph-mem MCP server (Python, local)
→ Graphiti REST API (Docker, httpx over HTTP)
→ Neo4j 5.26 (Docker)
→ OpenRouter API (LLM + embeddings)
- Language: Python, installed via
pip install -e . - No graphiti_core dependency: pure REST/HTTP client
- LLM backend: OpenRouter (configurable via
.env) - Embeddings: qwen/qwen3-embedding-8b (4096 native, truncated to 1024 client-side)
Docker Stack
docker compose up -d --build # from project root
Requires .env with OPENROUTER_API_KEY. See docker-compose.yml.
Patched files in graphiti/:
zep_graphiti.py— ExampleLLMClient (fixes schema echoing for Gemma/Qwen/Llama), configures embedder + cross-encoder from env varsingest.py— AsyncWorker catches exceptions instead of dying silently (upstream bug)Dockerfile— extendszepai/graphiti:latestwith patches
MCP Tools (current — minimal surface)
Only 3 tools exposed to keep model attention focused:
| Tool | Purpose |
|---|---|
__IMPORTANT__graph_mem |
Workflow reminder — tells model how to use graph-mem |
save_memory(content, scope) |
Save info. scope="user" or "project" |
search_memory(query, scope) |
Search. scope="all", "user", or "project" |
scope replaces the old group_id parameter — the server computes the correct group_id automatically from cwd's git remote URL. This prevents the model from inventing group_ids.
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.
- 3d ago First seen · 228 lines · 2,842 tokens per session scan B 1cc96d74cd0a
graph-mem CLAUDE.md is an instructions file published in the GitHub repository quequiere/graph-mem (5 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 2,842 tokens to every session, about $0.0142 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
Synaptic-Tuner CLAUDE.md
Instructions for ProfSynapse/Synaptic-Tuner, covering pact framework and managed project memory, pact routing, current session, retrieved context and pinned context.
claude-harness CLAUDE.md
Instructions for panayiotism/claude-harness, covering claude harness plugin, project overview, tech stack, session startup protocol and project structure.
seamless CLAUDE.md
Instructions for 0spoon/seamless, covering claude.md, what is seamless?, project structure, common commands and single test.
shinobi CLAUDE.md
Claude Code instructions for numbererikson/shinobi, covering claude.md, what this is, dogfooding — use shinobi to build shinobi, dev commands and layout.
agentboard AGENTS.md
AGENTS.md instructions for Vladev0/agentboard, covering agentboard — conventions for agents, quick start for a new agent, 1. primary path — mcp tools, project memory — a graph of knowledge notes and protocol for long loop/cycle sessions.
autospec CLAUDE.md
Claude Code instructions for Hundia/autospec, covering claude code memory — autospec, about this project, mandatory development workflow, rule 1: backlog-first development and rule 2: living documentation.