Gomaa
Local-first hierarchical memory engine giving AI agents persistent Obsidian-backed recall.
Tested on Linux
Use it
Linux x86_64
You need first
- Runtimes:
Python 3.9+
install gomaa from GitHub via pip install git+https://github.com/M4F-S/gomaa.git
pip install --user --upgrade git+https://github.com/M4F-S/gomaa.git
Where the README and reality differ
- README: pip install gomaa (or pip install "gomaa[all]")
What worked: Package not on PyPI; must install from git: pip install --user git+https://github.com/M4F-S/gomaa.git - README: uv tool install gomaa
What worked: Same PyPI missing error; requires git source
Check that it works
gomaa --help 2>&1The output should contain “Available commands”.
First thing to try
gomaa init 2>&1cat ~/.gomaa/vault/general/test/Test\ Memory\ Note.md 2>&1Add it to your agent
claude mcp add gomaa -- python3 -m gomaa serverJSON for Claude Desktop, Cursor and other clients
{
"mcpServers": {
"gomaa": {
"command": "python3",
"args": [
"-m",
"gomaa",
"server"
]
}
}
}Run gomaa init to set up the memory vault, then start the MCP server with python3 -m gomaa server and connect your agent's MCP client to it.
What it does
- Give AI agents memory that persistsTested
How to use it
- Run
gomaa initto set up the vault. - Connect your MCP client to
python3 -m gomaa server. - Call the memory_remember tool with a title, content, tags, wing and room.
- Run
- Search memories with hybrid RRF retrievalTested
How to use it
- Run
gomaa initand connect your MCP client topython3 -m gomaa server. - Call the memory_recall tool with your query, mode set to "hybrid", and a top_k value.
- Run
- Consolidate memories with a sleep cycleTested
How to use it
- Run
gomaa initand connect your MCP client topython3 -m gomaa server. - Call the memory_sleep tool, optionally setting decay_factor and archive_threshold.
- Run
- Store memories as Obsidian Markdown notesNot tested
How far pickfit got
Linux x86_64
- Installs
- Connects
- 3 of 4 features work
- Works offline
Good for
- store a private memory note in vault
- search memories across private and shared stores
- trigger sleep consolidation of episodes
- launch web knowledge graph dashboard
- ingest conversation transcript verbatim
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
An MCP client called one of its tools
- Replayed from scratch
- Works offline
- Checked on 31 Oct 2026
- Version 4.0.0
An MCP client handshake over stdio listed 14 tools, and a call to memory_remember returned a result.