What this enables
Capability record
Give agents local hybrid search, semantic retrieval, and shared memory collections without sending the indexed corpus to a remote embedding service.
SKILL / 1.0.0AS ABOVE / FIRST PARTY
Open curriculum / Applied capability
QMD Memory
Give agents local hybrid search, semantic retrieval, and shared memory collections without sending the indexed corpus to a remote embedding service.
memorylocal-firstsearchmcpmulti-agent
Capability map
What an agent can do
01
Search agent memory by keyword, meaning, or hybrid retrieval
02
Organize workspace, daily-log, intelligence, project, and document collections
03
Refresh local indexes on a schedule and share approved collections through MCP
Operating sequence
From source to useful action
01Scope
Choose which folders and records the agent is permitted to remember.
02Index
Build local keyword and vector indexes with collection context.
03Retrieve
Use hybrid search to return small, relevant, source-linked context.
04Maintain
Refresh the index and review collection boundaries as the workspace changes.
Implementation
Documentation
Install the public skill, run its setup workflow, choose only the collections the agent is allowed to index, and verify keyword, semantic, and hybrid retrieval before enabling shared MCP access.
Before you begin
- Node.js 22 or newer
- Approximately 3 GB of disk for local models and indexes
- A defined collection and retention boundary
Trust boundary
Use captured intelligence responsibly.
- Index only approved folders.
- Treat shared MCP access as a data boundary and expose only intended collections.
- Keep source citations attached so retrieved memory can be verified.