MemTether

Your AI agents' memory is a file, not a pipeline.

One physical SQLite database, shared by every AI client on your machine via file-level pointers. No cloud. No API fees. No sync. Your memory stays greppable, git-able, and yours.

Try it in your browser → · pip install · 中文文档

Quick Start

mcp-name: io.github.lanbass869-cell/memtether

# 1. Install
pip install memtether

# 2. Initialize (creates a demo memory DB)
memtether init

# 3. Connect all detected AI clients (Claude Code, Cursor, Windsurf, etc.)
memtether connect --all
memtether dashboard

Or connect one client at a time:

memtether setup claude-code   # writes MCP config + verifies
memtether setup cursor
memtether setup gemini-cli
# ... or run `memtether setup list` to see all 23 supported clients

# Search with L0 summaries (one line + uid, agent reads full separately)
memtether search "deploy" --list

This starts the API server and opens the web dashboard in your browser.

Try it:

memtether search "shared memory"
memtether remember "my first shared memory"
memtether stats

Or on Windows, double-click install.bat for one-click setup.

# From source
git clone https://github.com/MemTether/MemTether.git && cd MemTether && pip install -e .

# With semantic search (local embedding, no cloud)
pip install "memtether[vector]"

# With REST API server
pip install "memtether[server]"

# Docker
docker build -t memtether . && docker run -p 8420:8420 -v ./data:/app/data memtether
docker-compose up

python scripts/make_demo_db.py
MEM_DB=~/.memtether/demo.db python mem.py search "shared memory"

Why MemTether

The problem

Memory should be a data format, not a multi-stage pipeline. MemTether treats your agents’ shared memory as a plain SQLite file — ls it, grep it, git it, back it up. The symlink is the deploy step.

You use Claude Code for coding, Cursor for refactoring, and Windsurf for exploration. Each has its own memory. Switch tools and your AI forgets everything.

MemTether's answer is simpler than you'd expect: make them all point to the same file.

How it is different

Feature comparison

Architecture

+---------+   +---------+   +---------+   +---------+
|  Claude |   | Cursor  |   |Windsurf |   | VS Code |  ... 23 adapters
|  Code   |   |         |   |         |   |         |
+----+----+   +----+----+   +----+----+   +----+----+
     |              |              |              |
     +--------------+------+-------+--------------+
                          |
                   +------v------+
                   |  MemTether  |
                   | Memory Hub  |
                   |             |
                   | SQLite      |  <- one physical memory.db
                   | FTS5        |  <- full-text search (triggers)
                   | ChromaDB    |  <- vector search (bge-m3)
                   | Bi-temporal |  <- T + T-prime dual time axes
                   | Q-Value     |  <- usage-based ranking
                   | Hubguard    |  <- concurrent write lock
                   +-------------+

Connect Your Clients

python -m memtether connect --all
python -m memtether verify
python -m memtether detect
python -m memtether selftest

Supported (23): Claude Code, Cursor, Windsurf, VS Code, Zed, JetBrains, Cline, Roo Code, Kilo Code, Continue, Cody, Amazon Q, Gemini CLI, Neovim, Claude Desktop, Codex, WorkBuddy (CN + Intl), CodeBuddy, ZCode, DSH, OpenClaw, Agents-Neutral

Memory Operations

python -m memtether remember "User prefers dark theme" --source claude-code
python -m memtether search "theme preference"
python -m memtether correct <uid> "Updated text"
python -m memtether retire <uid> "No longer relevant"
python -m memtether stats
python -m memtether as-of 2026-09-15 --kind known
python -m memtether rebuild

Benchmarks

LongMemEval (500 questions, full run, 2026-10-02)

Results use our own harness. Not directly comparable with mem0 reported numbers.

Test suite

Known Limitations

1. tether_connect detect may falsely report "not connected" for same-source-different-path configs. Use verify for accurate results.
2. DSH cordis.patch.yml deep customizations cannot be safely rewritten. Use plan to preview.
3. MemTether shares one physical DB via file-level pointers. On Windows this uses NTFS junctions; on macOS/Linux, symlinks. Both are native OS features — no sync daemon needed.
4. Semantic search requires optional deps (chromadb, onnxruntime). Without them, degrades to keyword search.
5. Windows-first. macOS/Linux should work but not fully tested.

Relationship to Other Projects

• mem0: managed memory with cloud API. MemTether is for people who want everything local.
• cognee: knowledge graph + pipeline. MemTether is lightweight operational memory (SQLite, no Neo4j).
• letta (MemGPT): agent framework. MemTether works with existing agents you already use.
• engram: Go + SQLite + FTS5 + MCP. MemTether adds bi-temporal, source attribution, Q-Value, 23 adapters.

You can use MemTether alongside any of these.

REST API

MemTether includes a built-in REST API server (FastAPI) for non-CLI access:

# Start the API server
memtether dashboard --port 8820

A web dashboard is available at http://localhost:8820/dashboard when the API server is running.

License

Apache 2.0 - see LICENSE

Contributing

Issues and PRs welcome. See CONTRIBUTING.md.

Star History

中文文档 | Security | Changelog