LoreConvo — Claude Code MCP server: what it does, how to install
Local session memory for Claude and other agents, kept in SQLite with full-text search. Project context survives a restart.
claude mcp add loreconvo -e LORECONVO_PRO=<LORECONVO_PRO> -- uvx loreconvoWhat LoreConvo does
LoreConvo keeps your session history in ~/.loreconvo/sessions.db on your own machine. Two kinds of memory get written: what happened during a session (summaries, artifacts, questions left open) and which project decisions still stand. Search runs on FTS5, and queries containing identifiers such as autoSave or get_context_for are split into compound tokens first.
Inside Claude the server leans on hooks. auto_save.py writes a structured summary when a session ends; auto_load.py runs at the start of the next one, ranks recent sessions by how much signal they carry, and feeds the useful context back in. Sessions with open questions and decisions rank highest. A MEMORY.md in the project directory gets indexed too.
Cursor connects through .cursor/mcp.json, Codex through ~/.codex/config.toml, Gemini CLI through gemini mcp add. Automatic capture only happens if you install the session hooks yourself — otherwise you save memories by hand with the save tools. Nothing leaves your disk unless you turn on the optional summarization feature, which sends an excerpt to the Anthropic API with your own key and is off by default.
What it helps with
- pick up a project's context after restarting Claude
- track down which session covered a given module
- keep decisions from getting lost between sessions
- carry context from Claude over to Cursor or Codex
- copy the latest session to the clipboard for Claude Chat
Example prompts for the agent
- Save this session to project memory
- Remind me what we decided about the database schema
- Search past sessions for
pipeline_tracker
How to install LoreConvo
Platform CLI
claude mcp add loreconvo -e LORECONVO_PRO=<LORECONVO_PRO> -- uvx loreconvoReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from loreconvo <command>
Platform CLI
codex mcp add loreconvo --env LORECONVO_PRO=<LORECONVO_PRO> -- uvx loreconvoReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from loreconvo <command>
Config file
{
"mcpServers": {
"loreconvo": {
"command": "uvx",
"args": [
"loreconvo"
],
"env": {
"LORECONVO_PRO": "<LORECONVO_PRO>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from loreconvo <command>
Config file
{
"servers": {
"loreconvo": {
"type": "stdio",
"command": "uvx",
"args": [
"loreconvo"
],
"env": {
"LORECONVO_PRO": "<LORECONVO_PRO>"
}
}
}
}Replace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from loreconvo <command>
Platform CLI
gemini mcp add -e LORECONVO_PRO=<LORECONVO_PRO> loreconvo uvx loreconvoReplace <…> with your own values. uvx runs the command named after the package; if the README names another one, use uvx --from loreconvo <command>
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 LoreConvo
How do I install LoreConvo in Claude Code?
Run claude mcp add loreconvo -e LORECONVO_PRO=<LORECONVO_PRO> -- uvx loreconvo. The alternative is claude mcp add --env LORECONVO_PRO=<LORECONVO_PRO> --transport stdio loreconvo -- uvx loreconvo.
What does LoreConvo do?
It maintains a local SQLite store of session memory and exposes it to the agent over MCP: summaries, decisions, open questions. Lookups go through FTS5.
Where does LoreConvo store its data?
In ~/.loreconvo/sessions.db on your machine. There are no cloud accounts; summarization through the Anthropic API is a separate opt-in that needs your own key.
