code-session-memory: Search AI Coding Sessions with OpenAI Embeddings

Overview

code-session-memory is a TypeScript tool for automatically indexing and searching AI coding-agent sessions. It uses OpenAI embeddings, with text-embedding-3-large as the default model, to make conversations from OpenCode, Claude Code, Cursor, VS Code, Codex, and Gemini CLI semantically searchable through an MCP server, CLI, or web UI. Sessions from these tools share one database.

The default storage is a local SQLite database using sqlite-vec. An optional PostgreSQL and pgvector backend supports sharing indexed memory across machines. The project is aimed at developers who want an agent, or themselves, to retrieve useful decisions and solutions from earlier coding sessions without manually maintaining notes.

Key Features

• Automatically indexes new messages after an agent turn and skips previously indexed messages.
• Supports OpenCode, Claude Code, Cursor, VS Code, Codex, and Gemini CLI sessions in one database.
• Provides semantic session search through MCP tools and a command-line query interface, with source, project, and date filters.
• Includes an interactive CLI for browsing, printing, deleting, and compacting sessions into restart summaries.
• Offers a web dashboard for search, session browsing, status, and session and model analytics.
• Uses SQLite with sqlite-vec by default, or PostgreSQL with pgvector for centralized, multi-machine storage.
• Includes transcript parsers, analytics, and tests for the supported session sources.

Use Cases

• A developer using several coding agents can search prior conversations for an implementation decision without remembering which tool or session contained it.
• An agent can retrieve earlier solutions or project context through MCP when responding to questions about past work.
• A developer restarting work after a long session can generate a structured summary of context, decisions, progress, and unresolved issues.
• A team or individual working across multiple machines can configure PostgreSQL and pgvector to centralize indexed session memory.

Getting Started

Install the detected tool integrations with:

npx code-session-memory install

Set OPENAI_API_KEY, then restart the supported coding tools. See the README for environment setup, backend options, and usage details.

License in Plain Words

No standard license was detected. Without a license, the default copyright rules apply: you may read the code, but not reuse it. Check the repository or ask the authors.

Alternatives

• Agent-Memory: It captures and retrieves coding-agent session activity across similar tools, while code-session-memory indexes sessions with OpenAI embeddings in SQLite or PostgreSQL.
• a-mem-mcp: It builds a connected, self-evolving project-knowledge graph in ChromaDB and is tested with Claude Code, rather than indexing sessions across multiple agents.

Considerations

• Indexing and semantic queries require an OpenAI API key; the default embedding model is text-embedding-3-large.
• The project requires Node.js 22 or later. PostgreSQL is optional, but using it requires a configured PostgreSQL database with pgvector.
• The project is young, has no releases in the past year, and recent development is concentrated mostly in one contributor. Recent commits indicate ongoing work, but the small contributor base is relevant when assessing continuity and support.
• The README states an MIT license, while the repository metadata supplied for this analysis does not include a license field. Verify the repository’s license file before relying on its terms.
• Indexed conversation content is stored in the configured database. Consider what session data you want to retain and who can access a centralized backend before adopting it across machines.