Longmem
Give Claude Code or Cursor durable recall of decisions, APIs, and context across repos without re-pasting the same notes every session.
Overview
io.github.marerem/longmem is a MCP server for the Build phase that gives Cursor and Claude Code persistent cross-project memory via local semantic search.
What is this MCP server?
- Local persistent memory across projects with semantic search instead of brittle chat history
- stdio PyPI package (longmem) at version 0.2.1 for Claude Code and Cursor MCP config
- Cross-project recall so solo builders avoid repeating context dumps every new session
- Privacy-friendly local indexing versus sending full project history to a remote memory API
- Pairs with other dev MCP servers as the layer that remembers what those tools helped you decide
- Published package version 0.2.1 on PyPI identifier longmem
- stdio transport via MCP server schema 2025-12-11
- Targets Cursor and Claude Code per server description
What problem does it solve?
Coding agents forget everything between sessions and repos, so you waste time re-explaining stack choices, APIs, and half-finished decisions.
Who is it for?
Solo builders using Claude Code or Cursor on several codebases who want local, searchable agent memory without a third-party cloud notebook.
Skip if: Teams that need audited enterprise knowledge bases, strict RBAC, or zero local indexing on locked-down machines.
What do I get? / Deliverables
After you register longmem, your agent can semantically recall stored context across projects so implementation and fixes start from accumulated memory instead of a blank slate.
- MCP-accessible store-and-search memory across configured projects
- Semantic retrieval of past decisions and notes inside agent sessions
- Reduced repeated context pasting when switching repos or chats
Recommended MCP Servers
Journey fit
Agent memory is cataloged under Build because it is installed as MCP tooling that shapes how you code with assistants day to day. agent-tooling is the canonical shelf for MCP servers that extend what coding agents remember and retrieve, not a one-off integration.
How it compares
MCP memory integration with local semantic search, not a generic note-taking skill or hosted vector SaaS.
Common Questions / FAQ
Who is io.github.marerem/longmem for?
It is for indie developers and small teams on Cursor or Claude Code who want persistent, searchable context across multiple repositories without repeating long briefings.
When should I use io.github.marerem/longmem?
Use it once you rely on agents daily and notice the same project facts need to be re-stated after every chat reset or repo switch.
How do I add io.github.marerem/longmem to my agent?
Add the longmem PyPI package (0.2.1) as an stdio MCP server in your Claude Code or Cursor MCP settings, following the repository setup for your OS.