LoreConvo
labyrinth-analytics/loreconvo
Local SQLite memory for AI sessions that carries context across Claude Code, Cursor, Codex and Hermes
LoreConvo is a local, MCP-based memory layer that stores session summaries and decisions in a SQLite database you own. The README says the same memory is shared across Claude Code, Cowork, Codex, Cursor and Hermes Agent.
What LoreConvo does
LoreConvo keeps session memory in ~/.loreconvo/sessions.db, a file on your machine that you can inspect, edit, back up or delete. It records two kinds of memory per session: episodic, covering what happened, artifacts created and open questions, and semantic, covering stable decisions about the project. Each memory shows its provenance, including which surface captured it, when and under which project.
Capture is explicit through the save tools, or automatic at session end if you install the optional hooks. Claude Code gets all 39 LoreConvo MCP tools through its plugin loader. Search uses SQLite FTS5 with compound-token expansion for names like autoSave or pipeline_tracker. On a 60-session synthetic corpus with 36 labeled queries, the README reports Recall@5 of 88.9 percent and an MRR of 0.875, against 72.2 percent and 0.708 without the expansion. Optional AI summarization is a Pro feature, off by default, and sends a transcript excerpt to the Anthropic API with your own key.
Key features
- Cross-surface memory shared by Claude Code, Cowork, Codex, Cursor and Hermes Agent
- Local SQLite storage at ~/.loreconvo/sessions.db with full provenance per memory
- Episodic and semantic memory types
- 39 MCP tools and optional session hooks for automatic capture
- FTS5 search with camelCase and snake_case token expansion
- Published recall benchmark with a reproduction script
When to use it
- Carrying project context between Claude Code and Hermes Agent sessions
- Searching past decisions across tools by technical identifiers
- Auditing and deleting stored memories on your own machine
Who it is for: Developers who switch between several AI coding tools and want one local, inspectable memory.
How it fits with Hermes Agent
Supports Hermes Agent among several clients; the repository is tagged hermes and hermes-agent and exposes memory over MCP.
How to install LoreConvo
These commands are copied from the project's README. Check the repository for the latest steps before you run them.
uvx --from loreconvo[server] loreconvo
bash install.shNote: AI summarization is a Pro feature that is off by default and uses your own Anthropic API key.
FAQ
What is LoreConvo?
LoreConvo is a local memory layer for AI sessions. It stores summaries and decisions in a SQLite database you own and makes them available across Claude Code, Cowork, Codex, Cursor and Hermes Agent.
Does LoreConvo work with Hermes Agent?
Yes. The README names Hermes Agent among the supported clients that share the same memory, and the repository is tagged hermes-agent and mcp-server.
Is LoreConvo free and open source?
GitHub reports the license as NOASSERTION, meaning it could not identify a standard license. Read the LICENSE file and confirm terms with the author. Some features, such as AI summarization, are labeled Pro.
Similar memory for Hermes Agent
All memoryLocal-first project memory for AI coding agents, shared through plain files in .memory-bridge
Trainspotting31 MemoryBuddyShared memory server on Cloudflare Workers that Hermes and other MCP clients read and write
stepanov1975 Hermes Local KnowledgeHermes plugin that routes a local question to the skill, script, runbook or document to open first
aristoapp Membase Plugin for HermesPersistent long-term memory for Hermes Agent using hybrid vector search and a knowledge graph
alejandroiglesias Hermes DreamingBackground memory consolidation plugin that tidies Hermes MEMORY.md and USER.md on a schedule
haitao338241-collab Hermes Hindsight GuideChinese guide to a local Hermes memory setup with Ollama and Hindsight that reduces memory clutter
Related guides: SOUL.md for Hermes Agent: what it is and how to write one · Run multiple Hermes agents with profiles