MemSearch

MemSearch gives coding assistants searchable project history across sessions. It's for developers who want to recover earlier fixes, revisit design decisions or continue work in another agent.

• Shared agent memory: Plugins automatically capture conversations from Claude Code, Codex, DeepSeek Harness, OpenClaw and OpenCode, with the same project history searchable across them.
• Context retrieval: Search combines meaning and keyword matches, and you can expand results into full sections or original transcripts.
• Editable records: Markdown files hold the original memory, so you can review, edit and track it in Git independently of the rebuildable Milvus search index.
• Notes and reusable skills: Optional background tasks maintain project notes and user preferences or draft Agent Skills from recurring workflows; skill candidates stay inactive until you install them.

The default setup stores the index locally in Milvus Lite and runs ONNX bge-m3 embeddings on CPU without an API key. The embedding model requires an initial download. Ollama is another local embedding option. Teams can self-host Milvus Server with Docker, while Zilliz Cloud stores the index in a managed cloud service. Optional OpenAI, Google, Voyage, Jina and Mistral embedding providers send embedding requests outside your machine.

MemSearch is open source under MIT, with a CLI and Python API for custom agent integrations.