NLM Memory

NLM Memory collects coding conversations into one searchable store on your machine. It's for developers who switch agents or editors and need earlier reasoning to remain available when decisions change.

• Memory across tools: It indexes sessions from Claude Code, Codex CLI, Cursor, Windsurf, OpenCode, Hermes WebUI, Hermes Agent, Aider and pi.
• Editable history: Newer sessions can supersede earlier ones, and entities can be retired, so the timeline can distinguish current context from older reasoning.
• Agent recall: MCP gives compatible agents access to the shared index, with automatic context injection for Claude Code, Codex CLI, Hermes Agent and pi; Aider only contributes history.
• Search and review: A local browser interface provides keyword, semantic and hybrid search, a session timeline, conversation histories grouped by entity, and backup and restore controls.

The project is open source under Apache 2.0. Ollama handles classification and embeddings locally by default. Choosing DeepSeek, OpenAI, Anthropic, OpenRouter or another external OpenAI-compatible classifier sends session transcript content, including prompts and code snippets, to that provider. Provider API keys are stored in plaintext in the local SQLite database.

The service listens only on localhost. Access from another device requires a tunnel, such as Tailscale, rather than direct public exposure. It has no analytics SDK or crash reporter, but checks npm for update notifications unless disabled. Optional Telegram digests send their contents to Telegram.