Store and retrieve team knowledge in a PostgreSQL vector database so AI can learn from past decisions, solutions, and context across conversations. Built for teams that want institutional memory without manual documentation.
io.github.ysydhc/team-memory
Local install
STDIO
1 required env var
How models use it and what it is built for.
Store and retrieve team knowledge in a PostgreSQL vector database so AI can learn from past decisions, solutions, and context across conversations. Built for teams that want institutional memory without manual documentation.
Local install — runs as a subprocess.
Configuration this server reads at startup.
PostgreSQL connection string (with pgvector extension)
Path to config.yaml (optional)
Where to find authoritative docs and source for TeamMemory.
Paste any of these into Agent Studio after connecting TeamMemory.
Common questions about connecting and running TeamMemory.
What database does team-memory use?
Team Memory requires PostgreSQL with the pgvector extension enabled. You provide the connection string via the DATABASE_URL environment variable, allowing you to self-host or use a managed PostgreSQL service.
How do I set up team-memory?
Install via `uvx team_memory@0.1.2`, set the required DATABASE_URL environment variable to your PostgreSQL connection string, and optionally provide a TEAM_MEMORY_CONFIG path pointing to a config.yaml file for additional settings.
Can I customize how team memory stores or retrieves knowledge?
Yes, you can provide a config.yaml file by setting the TEAM_MEMORY_CONFIG environment variable. Refer to the project documentation for the schema and available configuration options.
Does team-memory use vector embeddings?
Yes, it uses pgvector (PostgreSQL's vector extension) to store and search embeddings, enabling semantic similarity search across team knowledge rather than just keyword matching.
What happens if I don't set DATABASE_URL?
The server will fail to start because DATABASE_URL is required. Make sure your PostgreSQL instance has pgvector installed and your connection string is valid before launching the MCP server.
MCP Playground runs 10,000+ hosted MCP servers — GitHub, Linear, Notion, Stripe, Sentry and more — across Claude, GPT, Gemini, DeepSeek and 30+ AI models. Compare model answers side-by-side, save agent presets, share runs. Zero install.
Open Agent Studio