Copilot Memory MCP
Persist and semantically search project memory locally with ChromaDB so your agent recalls decisions across long Build and Operate sessions without a cloud vector DB.
Overview
Copilot Memory MCP is a MCP server for the Build phase that provides persistent semantic memory for AI agents via local ChromaDB vector search without requiring cloud infrastructure.
What is this MCP server?
- Local ChromaDB vector search—no cloud memory service required
- Persistent semantic memory across agent sessions
- Docker OCI image `ghcr.io/labforgedev/copilot-memory-mcp:1.0.0` with SSE at localhost:8000
- Suited to coding copilots and long-horizon repo work
- Version 1.0.0 MCP server from LabForgeDev GitHub
- MCP server version 1.0.0
- Transport: SSE at http://localhost:8000/sse
- Backend: local ChromaDB vector search
Community signal: 1 GitHub stars.
What problem does it solve?
Long agent sessions forget prior decisions, so you re-explain stack choices and bug fixes every time you open the repo.
Who is it for?
Privacy-minded solo builders who want local-only agent memory on one machine with Docker already in the loop.
Skip if: Teams needing shared cross-machine memory, managed cloud recall, or memory without running a local container.
What do I get? / Deliverables
After you run the Docker SSE server locally, your agent can write and semantically recall project memory across sessions without a cloud vector database.
- Stored memory entries retrievable by semantic similarity
- Cross-session continuity for architecture and debugging notes
- Self-hosted memory plane with no required cloud vector DB
Recommended MCP Servers
Journey fit
Persistent agent memory is installed while you build agent workflows and stays valuable when you iterate in production—primary shelf remains Build agent-tooling on Skillselion. Agent-tooling fits because the server is local semantic memory (ChromaDB), not frontend UI, launch SEO, or finance data feeds.
How it compares
Local ChromaDB memory MCP, not a document OCR API or market-data finance integration.
Common Questions / FAQ
Who is Copilot Memory MCP for?
Developers using MCP-enabled coding agents who want persistent, semantically searchable project memory stored locally via ChromaDB.
When should I use Copilot Memory MCP?
Use it when you start wiring agent-tooling in Build and notice context loss across sessions, or while iterating in Operate on the same codebase and need durable recall.
How do I add Copilot Memory MCP to my agent?
Run the OCI image `ghcr.io/labforgedev/copilot-memory-mcp:1.0.0` (Docker), point your MCP client at the SSE URL `http://localhost:8000/sse`, then use the server’s memory store and search tools from your agent.