Back to directory

OpenCode-Mem0

An OpenCode plugin that gives coding agents persistent memory using a local SQLite and usearch vector database, auto-capturing facts from conversations and exposing search, profile and list operations as an agent tool, with no cloud service required.

OpenCode-Mem0

OpenCode-Mem0 adds persistent memory to the OpenCode coding-agent editor through a plugin that stores everything in a local SQLite database with the usearch vector extension; no cloud service is required and the plugin stays inert until explicitly activated. Memories are captured automatically from conversations by default, using a configurable memory provider (OpenAI, Anthropic or Google Gemini chat or embedding models) to extract what is worth remembering, or the plugin can reuse OpenCode's own main model and provider instead of a separate key.

A memory tool is exposed to the OpenCode agent with modes to add, search (hybrid vector plus full-text), read or write a user profile, and list recent memories, though the README notes most users will not need to call it directly since auto-capture and the web UI cover typical use. The plugin updates itself with a single command that also clears OpenCode's plugin cache.

It requires Bun 1.0 or later on Linux and macOS, or Node.js 20 or later on any platform including Windows.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.