codemem MCP server

codemem is a memory helper for coding assistants like Claude Code, OpenCode, and Codex. It quietly notes the decisions you make, the dead ends you hit, and the odd traps in your project, then reminds your AI about them later. It is handy if you keep forgetting why something was done a certain way.

What this MCP server does

You work as usual in your coding tool. codemem watches the session and writes down useful things, like what you decided and what did not work. When you start a new session and ask a question, your AI asks codemem for related memories. codemem searches its local store and hands back the relevant bits. Your AI then uses those bits to answer you with the right context.

Click to zoom

What you can do with it

• Recall a past decision and the reason behind it
• Find project-specific traps you ran into before
• Search your saved memories by keyword or meaning
• Browse memories and sessions in a local web viewer
• Share selected project memory with a teammate
• Sync memories across your own machines
• See what was captured and when

Try asking your AI

• “What decision did we make about the database migration, and why?”
• “Did we already try using a queue for this job? What happened?”
• “What is the known trap with the auth token refresh in this project?”
• “Show me recent memories about the payment service.”

What it gives back to you

You get answers in your chat that sound like your AI actually remembers the project. It can list related memories, summarize what was decided, or point to the original session. In the local viewer at localhost:38888 you can also browse the raw memories and sessions yourself. Some answers come from keyword search, some from meaning-based search, and codemem merges both.

Before you start

Who it's for

Developers and small teams who use AI coding assistants daily and want the assistant to remember project history across sessions.