Mcp Server
Persist lessons, memories, and semantic recall across coding sessions for Claude, Cursor, Copilot, and Windsurf.
Overview
Cachly MCP Server is an MCP server for the Build phase that provides persistent developer memory, lessons, and semantic search for AI coding agents.
What is this MCP server?
- Persistent memory and captured lessons across sessions
- Semantic search over stored dev context
- npm @cachly-dev/mcp-server v0.5.13 with stdio transport
- Positioned for Claude, Cursor, Copilot, and Windsurf
- Local-first AI dev brain rather than ephemeral chat history
- npm package @cachly-dev/mcp-server version 0.5.13
- Transport: stdio
- GitHub repository: cachly-dev/cachly
What problem does it solve?
Agents forget prior fixes, conventions, and lessons every new chat, so you repeat mistakes and re-onboard the model constantly.
Who is it for?
Long-horizon solo builders on multi-week repos who want cross-session agent recall without rebuilding prompts.
Skip if: One-off scripts or teams with strict policies against local AI memory stores.
What do I get? / Deliverables
Context and lessons persist locally and surface via semantic search so agents continue work with institutional memory intact.
- Persistent memory store queried by the agent during builds
- Semantic search hits over prior lessons and decisions
- Less repeated context-setting at session start
Recommended MCP Servers
Journey fit
A dev-brain MCP is installed while you build and refine agent workflows—canonical home is agent-tooling, with benefits into Operate iterate. Agent-tooling covers memory and semantic search layers that make long-horizon solo shipping possible.
How it compares
Persistent dev-memory MCP, not a public MCP directory or maps integration.
Common Questions / FAQ
Who is Cachly MCP Server for?
Indie developers using Claude, Cursor, Copilot, or Windsurf who need persistent lessons and semantic recall across coding sessions.
When should I use Cachly MCP Server?
Use it during Build agent-tooling when a project outgrows chat history, and keep it through Operate iterate when you revisit the same codebase.
How do I add Cachly MCP Server to my agent?
Install @cachly-dev/mcp-server from npm, configure stdio in your MCP client, and follow cachly-dev/cachly repository setup for memory storage paths.