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/biplav-crl/memroach/claude-mdgit clone --depth 1 https://github.com/biplav-crl/memroachWrote 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/biplav-crl/memroach/claude-md)<a href="https://agentmods.dev/instructions/biplav-crl/memroach/claude-md"><img src="https://agentmods.dev/badge/instructions/biplav-crl/memroach/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 | $0.00863 | $0.00863 |
| Opus 5 | $0.00432 | $0.00432 |
| Sonnet 5 | $0.00173 | $0.00173 |
| Haiku 4.5 | $0.00086 | $0.00086 |
Grade A, and why
memroach 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 4d 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.
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.
How it starts
The opening of the file, as written. The whole thing — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MemRoach — Instructions for Claude Code
What is this repo?
MemRoach is a CockroachDB-backed memory sync system for AI agents. It has three main components:
memroach_mcp_server.py— MCP server (primary interface, 16 tools)memroach_sync.py— File sync client + CLI + hook handler. Syncs~/.claude/to/from CockroachDB.memroach_daemon.py— Background sync daemon for real-time cross-machine updates.
Key patterns
- Single-file Python scripts using FastMCP and pg8000
- Direct CockroachDB connection (no Cloud Function intermediary)
- Content-addressable blob storage with gzip compression
- Hybrid search: vector embeddings (OpenAI text-embedding-3-small) + keyword matching
- Per-memory visibility:
private(default) orteam - Optimistic concurrency via version column
- Knowledge graph with typed links (relates_to, duplicates, supersedes, caused_by, refines)
- Memory decay via access tracking
Database
All tables prefixed with memroach_. Schema in schema/memroach_schema.sql.
memroach_blobs— deduplicated content storememroach_files— file metadata per user+machinememroach_embeddings— vector embeddings for searchmemroach_history— version changelogmemroach_links— knowledge graph edgesmemroach_access— read tracking for decay scoringmemroach_log— audit trail
Config
memroach_config.json (gitignored) — DB credentials, embedding API key, sync preferences.
How Claude Should Use MemRoach
Memory saving: files first, MCP for extras
Default behavior — no change needed. Continue saving memories to ~/.claude/memory/ files as normal. The sync hooks automatically push these to CockroachDB on Stop/SessionEnd. This is the primary path.
Use MCP tools when you need something files can't do:
memroach_store— Write a memory directly to DB (useful from non-Claude-Code clients like Cursor)memroach_share(path, "team")— Make a memory visible to teammatesmemroach_link(from, to, type)— Create knowledge graph connections between related memoriesmemroach_merge(paths, content)— Consolidate duplicate memories into one
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.
- 4d ago First seen · 80 lines · 863 tokens per session scan A d3aaacf3eda5
memroach CLAUDE.md is an instructions file published in the GitHub repository biplav-crl/memroach (1 stars, last pushed 5mo ago), licensed MIT. It adds 863 tokens to every session, about $0.0043 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.
Other instructions, from other repositories
automem CLAUDE.md
Instructions for verygoodplugins/automem, covering claude.md, project overview, development commands, setup environment and development.
cavemem CLAUDE.md
Instructions for JuliusBrussee/cavemem, covering cavemem — agent playbook, project identity, non-negotiable rules, architectural rules and layout.
recall CLAUDE.md
Instructions for joseairosa/recall, covering claude.md, project context, build & run commands, testing and unit tests with vitest (primary test runner).
memory-graph CLAUDE.md
Instructions for memory-graph/memory-graph, covering claude.md, master plan, build & test, architecture and key patterns.
mentedb copilot-instructions.md
Copilot instructions for nambok/mentedb, covering mentedb development instructions, project overview, workspace structure, build, test, and lint and key types.
remindb CLAUDE.md
Instructions for radimsem/remindb, covering claude.md, project at a glance, code map, where to read first — don't grep, don't ls and tread carefully — pause before implementing.