shodh-memory — Claude Code MCP server: what it does, how to install
Equips your agent with three-tier cognitive memory, vector semantic search, and a built-in GTD task manager.
claude mcp add --env SHODH_API_URL=<SHODH_API_URL> --env SHODH_API_KEY=<SHODH_API_KEY> --env SHODH_USER_ID=<SHODH_USER_ID> --transport stdio mcp-server -- npx -y @shodh/memory-mcpWhat shodh-memory does
The server organizes memory across working, session, and long-term stores following Cowan's cognitive framework. Vector lookups rely on MiniLM-L6 embeddings, while an internal knowledge graph tracks entities and strengthens frequent associations via Hebbian learning. Identical items are dropped automatically by checking content hashes.
Requests travel over standard HTTP by default, though you can route traffic through a Unix domain socket or a Windows named pipe via SHODH_IPC_ENDPOINT. The initial launch fetches roughly 38MB of dependencies, which include the native engine and embedding assets. Without a configured SHODH_API_KEY, the server refuses to run.
What it helps with
- retain context and technical decisions across separate agent sessions
- retrieve relevant project notes by meaning using vector embeddings
- track project tasks, subtasks, and reminders within GTD workflows
- prevent duplicate memories using automatic content-hash filtering
Example prompts for the agent
- Recall our earlier architectural decisions regarding the database schema.
- Save this code style preference into long-term memory with a styling tag.
- List all open tasks associated with the current refactoring milestone.
How to install shodh-memory
Platform CLI
claude mcp add --env SHODH_API_URL=<SHODH_API_URL> --env SHODH_API_KEY=<SHODH_API_KEY> --env SHODH_USER_ID=<SHODH_USER_ID> --transport stdio mcp-server -- npx -y @shodh/memory-mcpReplace <…> with your own values
Platform CLI
codex mcp add mcp-server --env SHODH_API_URL=<SHODH_API_URL> --env SHODH_API_KEY=<SHODH_API_KEY> --env SHODH_USER_ID=<SHODH_USER_ID> -- npx -y @shodh/memory-mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"mcp-server": {
"command": "npx",
"args": [
"-y",
"@shodh/memory-mcp"
],
"env": {
"SHODH_API_URL": "<SHODH_API_URL>",
"SHODH_API_KEY": "<SHODH_API_KEY>",
"SHODH_USER_ID": "<SHODH_USER_ID>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"mcp-server": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@shodh/memory-mcp"
],
"env": {
"SHODH_API_URL": "<SHODH_API_URL>",
"SHODH_API_KEY": "<SHODH_API_KEY>",
"SHODH_USER_ID": "<SHODH_USER_ID>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e SHODH_API_URL=<SHODH_API_URL> -e SHODH_API_KEY=<SHODH_API_KEY> -e SHODH_USER_ID=<SHODH_USER_ID> mcp-server npx -y @shodh/memory-mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about shodh-memory
How do I install shodh-memory in Claude Code?
Run claude mcp add shodh-memory -e SHODH_API_KEY=<SHODH_API_KEY> -- npx -y @shodh/memory-mcp, substituting <SHODH_API_KEY> with your actual credential.
Does shodh-memory require an internet connection after setup?
Connection is only necessary on the first start to download the engine binary and MiniLM-L6 files totaling approximately 38MB. Subsequent memory lookups run offline.
What tools does this server provide to the model?
It exposes 51 tools, including remember for storing facts, recall for vector lookups, and proactive_context to pull relevant records into ongoing discussions.
