brainvault CLAUDE.md

Project instructions for Brainvault, a local memory system that stores information in SQLite, a small file-based database. They describe its structure, supported coding tools, installation limits, and how its command-line and integration files work.

In plain words
What is it for?
Use them when developing, installing, or running Brainvault, including its database, command-line interface, memory capture, or connections to Claude Code and Cursor.
Why use it?
They give coding agents the rules and background needed to work safely in the project. They also prevent assumptions about external services, Docker, or unsupported setups.

Instructions file

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/sumithsb/brainvault/claude-md
Clone the repo
git clone --depth 1 https://github.com/SumithSB/brainvault
Per session 1,898 This file is loaded in full into every session.
When invoked 1,898 The same file — it is already loaded in full.
Security scan B 1 finding. 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.

ModelPer sessionOnce invoked
Fable 5 $0.01898 $0.01898
Opus 5 $0.00949 $0.00949
Sonnet 5 $0.00380 $0.00380
Haiku 4.5 $0.00190 $0.00190

Measured 2d ago against content hash d2201a6670e0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade B, and why

brainvault 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 2d 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.

- `claude_code.py` — `~/.claude/settings.json` (MCP + Stop + PostToolUse hooks) + `~/.claude/CLAUDE.md` + transcript parsing
CLAUDE.md · 112 lines

How it starts

The opening of the file, as written. The whole thing — 112 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Brainvault

Personal memory layer for Claude Code and Cursor. SQLite + FTS5 + optional semantic search + MCP. Zero infrastructure.

Key constraints

  • No Docker, no external services, SQLite only
  • Python 3.10+, supports Claude Code + Cursor (one install targets every detected host)
  • Optional semantic extras: pip install 'brainvault[semantic]' (fastembed + sqlite-vec)

Structure

  • brainvault/db.py — storage layer: SQLite schema, FTS5, vector search, reflection queries
  • brainvault/mcp_server.py — 12 MCP tools served via stdio
  • brainvault/cli.py — CLI entry point (23 commands)
  • brainvault/capture.py — Stop hook entry (maintenance + session save); iterates adapters for recent transcript JSONL
  • brainvault/tool_capture.py — Tool-hook stdin reader; dispatches payloads to adapters (owns_payload / event_from_payload); <20 ms per event, never crashes
  • brainvault/adapters/ — per-host integration layer.
    • base.pyAgentAdapter ABC + HookResult / SessionEvent dataclasses
    • claude_code.py~/.claude/settings.json (MCP + Stop + PostToolUse hooks) + ~/.claude/CLAUDE.md + transcript parsing
    • cursor.py~/.cursor/mcp.json + ~/.cursor/rules/brainvault.mdc + ~/.cursor/hooks.json (stop + postToolUse + afterFileEdit + afterShellExecution) + ~/.cursor/projects/*/agent-transcripts/*/*.jsonl
  • brainvault/installer.py — dispatches install / uninstall over every detected adapter; auto-seeds only when the vault has zero memories (otherwise skip with a hint)
  • brainvault/bootstrap.py — bulk-imports session transcripts (Claude Code ~/.claude/projects + Cursor ~/.cursor/projects/.../agent-transcripts)
  • brainvault/git_scan.py — mines git history for architectural decision memories; discover_repos() for full-system scan
  • brainvault/code_scan.py — file tree walker, regex import extractor, co-change matrix builder; index_repo() orchestrator
  • brainvault/embeddings.py — fastembed wrapper (BAAI/bge-small-en-v1.5, lazy-loaded singleton)
  • brainvault/graph.py — generates self-contained HTML brain graph (D3.js force-directed); two-layer visualization: memory circles + teal diamond code file nodes (from code_entities/ code_cochange tables populated by index-repo); 6 edge types: belongs_to, file_overlap, temporal, keyword_overlap, cochange (files that change together), memory_file (commit→file); Layer filter chips, edge toggles, full-text search across paths/languages/keywords

Read the full file on GitHub · 112 lines

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.

  1. 2d ago First seen · 112 lines · 1,898 tokens per session scan B d2201a6670e0

Subscribe to this mod's changes

brainvault CLAUDE.md is an instructions file published in the GitHub repository SumithSB/brainvault (2 stars, last pushed 2mo ago), licensed MIT. It adds 1,898 tokens to every session, about $0.0095 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.

Related

Other instructions, from other repositories