Local-first long-term memory for AI agents - plain Markdown vault + thin MCP server. CJK-aware BM25, token-efficient by design.
Local-first long-term memory for AI agents — a plain Markdown vault + a thin MCP server. 给 AI Agent 用的本地长期记忆：纯 Markdown 知识库 + 薄 MCP server。
把上面 JSON（vault 路径换成你的）粘进任意 MCP 客户端（Claude Code / Codex / Cursor / DSH / Open WebUI…），重启客户端，完成。Agent 从此有了跨会话、跨工具的长期记忆。
Paste that JSON (with your vault path) into any MCP client and restart it — done. Your agents now share one long-term memory.
- Plain Markdown, no lock-in — your memory is a folder of .md files. Open it in Obsidian, grep it, version it with Git. Remove agentbrain and the memory stays. - Token-efficient by design — index-first retrieval: Index.md is the cheap first layer, BM25 (CJK-aware) only ranks candidates, and query output is compact by default (mode='index'); full text only on demand. - Append-only for agents — agents may create lessons, never edit or delete them. Consolidation happens through proposals in consolidations/ that a human approves, which keeps multi-agent writes conflict-free. - Plug-and-play via MCP — one server, every client: Claude Code, Codex CLI, OpenCode, Cursor, DSH, Open WebUI, ... - Secrets never enter the vault — credentials live in env/keyring; lessons reference ${ENV:VARNAME} placeholders only, resolved at runtime via shell. Since 0.4.1 this is enforced, not just a rule: memoryingest scans for credential-shaped content (sk-/ghp/AKIA/xox-/AIza keys, bearer tokens, private-key blocks, password= assignments) and refuses the write, telling the agent to use a placeholder instead. Placeholders and teaching examples (sk-xxx, YOURKEY) ingest fine. lint also scans existing lessons and reports SECR
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/2672243194--agentbrain.html)
Open plain-text file format for AI memory. Your assistant's long-term memory as .dai files on your disk: readable by Claude, GPT, Gemini, Cursor, local models and grep (all LLM models work). MCP serve
rwnalds/engramSelf-hosted MCP server + dashboard giving Claude Code, Cursor & Hermes agents shared long-term memory over a git-backed markdown vault. Agent-native, Obsidian-compatible second brain — markdown RAG, n
osaxyz/sepiaceLong-term memory for coding agents. The sepiace plugin for Claude Code and Codex.
holetexvn/memgwShared long-term memory for AI coding agents — one local store for Claude Code, Codex CLI, opencode and anything that speaks MCP. SQLite + FTS5, no vector DB, $1-3/month.
tinhien11/remem-mcpLong-term memory MCP server for AI coding agents. SQLite + FTS5 + vector search, CodeGraph, error learning, compaction survival. Beats Mem0 on benchmarks. No cloud, no API key.
cmoraes10/lociTyped long-term memory for AI agents. One core, an MCP server, and a Hermes plugin.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.