Serveur MCP Renor | Awesome MCP Servers

Your AI should remember.

Renor is one memory for all your coding agents. Tell one agent once, and every agent can recall it, in any session.

Get started See how it works

Works with Claude Code Codex OpenCode Pi

Copy setup prompt

Paste it into your agent. It connects Renor, then you sign in once in the browser. No key to copy.

Cursor, ChatGPT, Claude.ai and others: see the Quickstart.

Tell it once.

Every agent session starts from zero. Renor is a memory layer your agents share over MCP: they save what matters as you work and look it up before they answer.

An example

An example: on Monday you tell Claude Code “Use pnpm in this repo, never npm.” and it saves that to Renor. On Thursday Codex recalls it before adding a package and runs pnpm. Two weeks later OpenCode recalls it while setting up CI.

Tell one agent once and the others can recall it: Claude Code, Codex, Cursor, OpenCode, Pi, ChatGPT and Claude.ai.

Yours to see. Yours to take.

Everything your agents save shows up in your dashboard. Search it, see how it changed, and export all of it as plain Markdown any time.

The Renor dashboard. Sample data.

How it works.

Set up once. After that it works inside the agents you already use, with nothing new to learn.

01 Connect Paste the setup prompt into your agent. It adds Renor as an MCP server and installs the Renor skill.
02 Remember As you work, your agent saves durable facts, decisions and preferences. Or just say “remember that…”.
03 Recall Before it answers, your agent pulls the relevant context from Renor. You stop repeating yourself.

Measured, not claimed.

LongMemEval-S tests whether a memory system can answer questions about long, multi-session conversation histories.

361 of 500 LongMemEval-S questions answered correctly with Renor as the memory. That is 72.2%, the product score of record.

Recorded 2026-09-22 with the shipped defaults and a strict direct-answer prompt. A different setup scored 86.4% (432 of 500) on 2026-09-30, with an evidence model a new workspace does not run, so it is not the score of record.

Methodology and full results