Skip to the collection

agent-native · Memory & State · Verified 2026-10-13

MemSearch

Listed on The Agent-Native Index

Cross-platform semantic memory for AI coding agents

What it does

MemSearch is Zilliz’s search-over-memory layer for coding agents. Official GitHub description: “A persistent, unified memory layer for all your AI agents (e.g. Claude Code, Codex, DSH), backed by Markdown and Milvus.” Docs: install a plugin and the agent remembers prior work across sessions — conversations are captured, indexed with hybrid search, and recalled when needed.

Markdown under .memsearch/memory/ is the source of truth. Milvus is a derived, rebuildable index (dense + BM25 + RRF). Recall is progressive: search summaries, expand a section, then drill into the original transcript. The same collection-naming algorithm lets Claude Code, Codex, DeepSeek Harness, OpenClaw, and OpenCode share one project memory.

Official positioning is retrieval over agent memory, not a general fact-extraction memory OS. That is the distinction versus Mem0, Zep, and Memoria.


Agent-facing interfaces

How agents onboard

Interaction pattern: CLI + platform plugins

# CLI / library (ONNX embeddings, no API key)
uv tool install "memsearch[onnx]"

memsearch index ./memory/
memsearch search "how to configure Redis?" --top-k 5

See the full onboarding section in the dossier.

Protocol surface

Interface Detail
CLI memsearch index, search, expand, watch
Python API MemSearch(paths=...).index() / .search()
Claude Code plugin Marketplace zilliztech/memsearch
DSH plugin dsh plugin --profile web add @zilliz/memsearch-dsh
OpenClaw openclaw plugins install --force clawhub:memsearch
Storage Markdown + Milvus Lite / Server / Zilliz Cloud

Verification

Last verified 2026-10-13.

Sources: