Shared Memory for Devin Desktop

Shared with your team and across your AI tools.

MMRY AI gives Devin Desktop, the new name for Windsurf, shared memory that you share with your team and across your AI tools: one memory that Devin Desktop, Cursor, Claude, ChatGPT, Claude Code and your other connected AI tools all read and save. Add MMRY AI to Devin Desktop as a remote MCP server once, and what you save there is there in Claude Code, Cursor and ChatGPT when they are connected to the same account. Memories you share with your team are there for each teammate in the AI tools they connect.

Start your free trial

14-day free trial.

How to set up MMRY AI in Devin Desktop

  1. Start your free trial. Create your MMRY AI account at mmryai.com/setup.
  2. Open your MCP config file. Open mcp_config.json in your Devin settings folder: ~/.config/devin/mcp_config.json on macOS and Linux, or %APPDATA%\devin\mcp_config.json on Windows. If the file is not there yet, create it.
  3. Add MMRY AI. Under mcpServers, add an entry named mmry with the url https://mmryai.com/mcp, as in the example below these steps, and save the file.
  4. Sign in. When the MMRY AI sign-in page opens in your browser, sign in to your MMRY AI account and approve the connection. If Devin Desktop shows MMRY AI as Needs auth in the MCP list, select Authenticate to open that page.
  5. Use it in a conversation. Start a new conversation, ask the agent to load your MMRY memories, and allow the MMRY AI tool when Devin Desktop asks. You can allow every tool on the MMRY AI server, for the session or permanently.

The MMRY AI entry in mcp_config.json: { "mcpServers": { "mmry": { "url": "https://mmryai.com/mcp" } } }

If the file already lists other servers, add the mmry entry inside mcpServers next to them. Devin documents this file for both the Devin Local agent, the default for new conversations, and the legacy Cascade agent.

Share the setup with your team

To give everyone on a project the same setup, add the same entry to .devin/mcp_config.json in the project folder and commit it; Devin documents project servers as shared with your team through version control. Each teammate signs in to their own MMRY AI account.

Team admins can limit which MCP servers people may use with an allowlist or an MCP registry. If yours does, ask your admin to allow MMRY AI; on the allowlist, the server ID must match the entry name mmry.

What works in Devin Desktop

Shared memory for Devin Desktop: FAQ

What is shared memory for Devin Desktop?

Shared memory for Devin Desktop means memory shared with your team and across your AI tools. Devin Desktop reads and saves the same memories as your other AI tools, and the memories you share with your team are there for each teammate in the AI tools they connect. MMRY AI keeps those memories in your MMRY AI account, so what you save in Devin Desktop is there in Claude Code, Cursor, Claude, ChatGPT and Codex when they are connected to the same account, and the other way round.

Is Devin Desktop the same as Windsurf?

Yes. Devin Desktop is the new name for Windsurf, and it arrives as an update with your plan and settings carried over. Set up MMRY AI in it with the steps on this page.

Does Devin Desktop remember previous conversations with MMRY AI?

Yes, through your MMRY AI account. In a new conversation, ask the agent to load your MMRY memories and it picks up your most recent memories, or ask it to search your MMRY memories for something specific.

How do I add a remote MCP server to Devin Desktop?

Add an entry named mmry with the url https://mmryai.com/mcp under mcpServers in mcp_config.json in your Devin settings folder, and save the file. Then sign in to your MMRY AI account when the sign-in page opens, and approve the connection.

Do I need a client ID or secret to connect Devin Desktop?

No. MMRY AI supports dynamic client registration, so Devin registers itself and you only sign in. A fixed OAuth client ID is only for servers that do not support dynamic client registration.

Which Devin Desktop plans can use MMRY AI?

Devin names no plan requirement for MCP servers in its MCP documentation. On Enterprise plans, MCP for the Cascade agent is turned on in settings, an admin turns on the Devin Local agent for members, and admins can limit MCP servers with an allowlist.

Can my team share memories through Devin Desktop?

Yes. Each person signs in to their own MMRY AI account in your organization and connects Devin Desktop, and each memory is private, shared with a memory group, or open to your organization. To give everyone on a project the same setup, add MMRY AI to .devin/mcp_config.json in the project folder and commit it.

Can Devin Desktop share memory with Claude Code and Cursor?

Yes. Connect Devin Desktop, Claude Code and Cursor to the same MMRY AI account, and they read and save the same memories. Cursor, Claude and ChatGPT connect with the same address, https://mmryai.com/mcp, and Claude Code and Codex connect through the MMRY AI plugin.

How do I correct or remove a memory I saved in Devin Desktop?

Ask the agent to save the correction as a new memory, and remove a memory in Claude Code or Codex. The agent can also make a memory you saved private, so only you see it.

Is there a free trial?

Yes. MMRY AI starts with a 14-day free trial, then continues on a per-user monthly plan; the current price is at mmryai.com/pricing. Every user adds credits to your team's shared pool, and if the pool runs out before the month ends your plan renews early so your team keeps working.

See all AI Platforms MMRY AI works with

Give Devin Desktop memory shared with your team and your AI tools

Start your free trial, connect Devin Desktop, and use the same memories in every tool you connect.

Start your free trial

14-day free trial.