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/vimo-ai/memex/claude-mdgit clone --depth 1 https://github.com/vimo-ai/memexWrote 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/vimo-ai/memex/claude-md)<a href="https://agentmods.dev/instructions/vimo-ai/memex/claude-md"><img src="https://agentmods.dev/badge/instructions/vimo-ai/memex/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 | $0.01000 | $0.01000 |
| Opus 5 | $0.00500 | $0.00500 |
| Sonnet 5 | $0.00200 | $0.00200 |
| Haiku 4.5 | $0.00100 | $0.00100 |
Grade A, and why
memex 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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
项目简介
Memex 是 Claude Code 会话历史管理系统（Rust 实现），解决本地数据 30 天过期、历史对话难检索、知识无沉淀的问题。
架构
V2 读写分离：文件监听和写入由 vimo-agent 统一负责，memex-rs 通过 HTTP API 被动触发 compact 和向量索引。
两种运行模式：
- CLI 模式（默认）：本地运行，含文件监听、定时任务、Web UI、MCP
- Server 模式：接收客户端 push，提供跨人搜索，不跑 LLM/embedding
技术栈
- 语言: Rust
- HTTP: Axum + tower-http
- 存储: SQLite + FTS5（文本）、LanceDB（向量）
- Embedding: BGE-M3（Ollama 本地，1024 维）
- LLM: Ollama（compact 用 Qwen3 0.6B，knowledge 用强模型）
- 协议: MCP（JSON-RPC 2.0）、HTTP API
- 共享依赖: ai-cli-session-db（reader + search + sync）
核心模块
| 模块 | 说明 |
|---|---|
compact/ |
LLM 多层压缩：L0 原文 → L1 Observations → L2 Talk Summary → L3 Session Summary |
knowledge/ |
L4 知识结晶：从 L0 提取结构化知识（集群/节点/关系），用强模型，独立于 compact |
search/ |
混合检索：FTS5 + 向量搜索 + RRF 融合，支持 L0-L3 多级别 |
vector/ |
LanceDB 向量存储 |
embedding/ |
BGE-M3 embedding |
rag/ |
RAG 问答：混合检索 → 上下文构建 → LLM 生成 |
mcp/ |
MCP 工具：search_history / get_session / get_recent_sessions / list_projects |
server/ |
Server 模式：ingest / register / search |
db_reader/ |
只读数据库层（替代 SharedDbAdapter） |
api/ |
HTTP API 路由 |
backup/ |
增量备份 |
archive/ |
分层归档（daily → weekly → monthly → yearly） |
indexer/ |
向量索引器 |
inject/ |
上下文注入 |
Compact 设计
原文始终保留，压缩层可选、可重做：
- L0: 原文（messages 表）
- L1: Observations — 每个工具调用一个（0.6B 模型）
- L2: Talk Summary — 每轮对话一个（0.6B 模型）
- L3: Session Summary — 整个会话一个，依赖 L2（0.6B 模型）
- L4: Knowledge — 结构化知识提取（强模型，独立管线）
MCP 设计原则
- 精简输出：只返回 AI 需要的字段，减少 token 消耗
- 渐进披露：列表返回摘要，详情返回完整内容
- 位置导航：
at字段定位消息，around模式获取上下文
关键路径
| 路径 | 说明 |
|---|---|
memex-rs/ |
Rust 主项目 |
memex-lite/ |
轻量版 |
docker/ |
Docker 部署配置 |
web/ |
Web UI |
~/.vimo/db/ |
数据库文件 |
~/.claude/projects/ |
Claude Code 会话数据源 |
环境变量
PORT— 服务端口（默认 10013）MEMEX_DATA_DIR— 数据目录（默认 ~/.vimo/db）MEMEX_WEB_DIR— Web 静态文件（默认 ~/.vimo/memex/web）OLLAMA_API— Ollama API URL（默认 http://localhost:11434）
开发约定
- 中文注释和文档
- 先实现核心功能，不过度设计
- compact 模块用 0.6B 小模型保持低成本，knowledge 用强模型保证质量
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 · 82 lines · 1,000 tokens per session scan A 309fde76e1af
memex CLAUDE.md is an instructions file published in the GitHub repository vimo-ai/memex (28 stars, last pushed 2mo ago), licensed MIT. It adds 1,000 tokens to every session, about $0.0050 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 2026-08-30.
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).
mentedb copilot-instructions.md
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.
mentisdb AGENTS.md
Instructions for CloudLLM-ai/mentisdb, covering mentisdb standards, scope, standards and design bias.