Decrec MCP server

What it does

DecRec is a decision memory system that records consequential choices made during AI agent conversations and makes them available across different AI tools and future sessions. It lets you record a decision once (e.g., in ChatGPT) and have other agents (e.g., Cursor) recall and propose revisiting it weeks later.

• Record a design or architecture decision made in one conversation, then have a different AI agent surface that decision and its reasoning when you later consider changing course.
• Keep a shared record of product or technical choices across team members who work with different AI tools (Claude, ChatGPT, Cursor, Codex).
• Surface past decision context to a fresh agent without manually restating the reasoning, constraints, and alternatives that led to the original choice.

Tools this server exposes

Tool	Does
search_decisions	Search recorded decisions by query or criteria.
create_decision	Prepare a decision record for human review and commitment.
update_decision	Modify an existing decision record.
commit_decision	Finalize and persist a decision to shared memory.
archive_decision	Mark a decision as no longer active or relevant.
admit_decision	Accept and acknowledge a prepared decision record.

Before you enable it

An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.