memoweft — Claude Code MCP server: AI memory with source tracking
memoweft stores AI conversation memory with structure: raw facts separate from derived understanding, with corrections, retractions and access control.
MIT ★ 6 updated 2026-10-24
claude mcp add memoweft -e MEMOWEFT_DB_PATH=<MEMOWEFT_DB_PATH> -- npx -y @memoweft/mcp-serverWhat memoweft does
memoweft records what the user said as Evidence — verbatim text tagged with provenance. From these records the engine builds World objects: entities, relationships, events, evaluations, cognitions. When the user corrects the AI or retracts a fact, no version is lost — the system knows the origin of each piece.
The server exposes eight tools: five for reading (search, history, dependencies), three for writing (new fact, correction, deletion). Forgetting is selective: the engine previews what else depends on that fact and removes linked copies in conversations. Requires MEMOWEFT_DB_PATH environment variable pointing to a local SQLite database. Node 24 uses the built-in driver; Node 20 and 22 pull in better-sqlite3.
What it helps with
- track where the AI got every remembered fact
- forget a fact along with dependent mentions in conversations
- store a user correction and show the original version
- separate verbatim user statements from model interpretation
- maintain a log of memory retractions and corrections
Example prompts for the agent
- Remember the project uses PostgreSQL 15 with the
pgvectorextension - Show all facts about me from the last three conversations
- Forget what I said about migrating to TypeScript
How to install memoweft
Platform CLI
claude mcp add memoweft -e MEMOWEFT_DB_PATH=<MEMOWEFT_DB_PATH> -- npx -y @memoweft/mcp-serverReplace <…> with your own values
Platform CLI
codex mcp add memoweft --env MEMOWEFT_DB_PATH=<MEMOWEFT_DB_PATH> -- npx -y @memoweft/mcp-serverReplace <…> with your own values
Config file
{
"mcpServers": {
"memoweft": {
"command": "npx",
"args": [
"-y",
"@memoweft/mcp-server"
],
"env": {
"MEMOWEFT_DB_PATH": "<MEMOWEFT_DB_PATH>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"memoweft": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@memoweft/mcp-server"
],
"env": {
"MEMOWEFT_DB_PATH": "<MEMOWEFT_DB_PATH>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e MEMOWEFT_DB_PATH=<MEMOWEFT_DB_PATH> memoweft npx -y @memoweft/mcp-serverReplace <…> 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 memoweft
How to install memoweft in Claude Code?
Run claude mcp add memoweft -e MEMOWEFT_DB_PATH=<MEMOWEFT_DB_PATH> -- npx -y @memoweft/mcp-server, where <MEMOWEFT_DB_PATH> is the path to the SQLite database file on disk.
What does memoweft do?
The server records AI conversation memory with separation between raw user statements (Evidence) and derived objects (Entity, Relationship, Event). Every fact is stored with its source, corrections and dependencies, enabling selective forgetting and explaining where a memory came from.
Which Node version does memoweft need?
Node 24 works with built-in SQLite; Node 20 and 22 require installing the optional better-sqlite3 dependency.