Persistent memory for Codex

Connect Codex to Sirro to save useful work once and reuse the same source in another repository.

Why coding agents need memory

Valuable work often dies in chat logs. Coding agents forget decisions, code, and reusable components between sessions and projects, so developers rebuild work that already exists.

How Sirro solves it

Save assets deliberately. Choose the code, component, animation, prompt, pattern, or link worth keeping and give it a clear name.
Retrieve across projects and agents. Ask a connected coding agent for the saved asset by name from another repository or session.
MCP native. Sirro exposes save, retrieve, list, compose, and update tools through its remote MCP server.

Codex MCP config

Add the remote Sirro MCP server. Your coding client opens Sirro so you can sign in and allow access to a library.

[mcp_servers.sirro]
url = "https://www.sirro.dev/mcp"

Sirro closed beta

Sirro is currently in closed beta. Open the beta access page.

Frequently asked questions

Can Codex retrieve an asset without searching old projects?

Yes. Ask Codex to use a named asset from Sirro. The Sirro MCP tool returns the stored asset directly from the connected library.

What can Codex save to Sirro?

Codex can save reusable code, components, animations, prompts, patterns, and links. Complete code saves can contain several files and their required styles and dependencies.

Does Codex memory work across repositories?

Yes. Sirro is separate from the current repository, so Codex can retrieve the same named asset while working in another codebase.

Does Codex need an API key in its config?

No. The remote MCP connection uses browser sign in and approval. The config only contains the Sirro MCP URL.

Sirro home, Shared memory for Cursor, Memory for Claude Code across projects, External memory for AI coding agents, Shared context for coding agents, An MCP memory server for coding agents.