How to Configure Windsurf SSH MCP Server

Published
Length
3 min read

Luis Felipe Dias

All posts

To set up a windsurf ssh mcp server, add the server entry in the editor settings. Then give the local tool a valid key or certificate for your remote Linux host. Note that Cognition renamed Windsurf to Devin Desktop in June 2026. The editor keeps the same features and still uses the Cascade agent. Teams often set up this agent with broad Linux accounts and static keys. When permissions are too wide, the agent can run bad commands or read private files. How a windsurf ssh mcp server works in pr

SSH Bastion Hosts / Jump Servers + Kubernetes API Server Access: the complete guide

Architecture patterns, implementation strategies, and security best practices. Delivered to your inbox.

Free white paper. No spam. Unsubscribe anytime.

To set up a windsurf ssh mcp server, add the server entry in the editor settings. Then give the local tool a valid key or certificate for your remote Linux host. Note that Cognition renamed Windsurf to Devin Desktop in June 2026. The editor keeps the same features and still uses the Cascade agent. Teams often set up this agent with broad Linux accounts and static keys. When permissions are too wide, the agent can run bad commands or read private files.

How a windsurf ssh mcp server works in practice

The Cascade agent inside the editor links to local Model Context Protocol servers. It reads the local server settings file to find available tools. The agent can also run terminal commands on its own after the user approves. It opens a dedicated shell that loads the user profile. On macOS it uses zsh, and on Linux it uses bash. Because the shell loads user settings, local aliases and variables stay active.

When the agent calls the SSH tool, the local server opens a network session. It reaches the target host using standard OpenSSH keys or signed certificates. OpenSSH added certificate support for users and hosts in version 5.4. A certificate authority key signs a public key to build a user certificate. This certificate holds the public key, identity info, valid user names, and options. It uses a much simpler format than standard TLS certificates.

Once the remote daemon checks the key or certificate, it starts a shell. The MCP server sends the shell command from the AI agent to the host. The remote host runs the command and sends the text output back. The agent reads that output in the chat window and plans its next step.

How to set up OpenSSH keys and roles for the agent

A safe setup starts with a dedicated Linux user on the remote host. Do not give the AI agent root access or full sudo rights. Instead, create a role account that can only read specific log folders or app files.

Next, use OpenSSH user certificates instead of static private keys on the laptop. When you sign a key, you set a key identity and list allowed user names. You should also set a short valid time window on the certificate. That way, the local file expires on its own after a few hours. If a laptop gets lost, an old key will not let anyone log in.

Still, you must store that key or certificate on the developer machine. The local MCP server needs to read the file to log in to the host. Any other app or script running under that local user account can also read it.

SSH Bastion Hosts / Jump Servers + Kubernetes API Server Access: architecture patterns and best practices

Continue reading? Get the full guide.

Free. No spam. Unsubscribe anytime.

What native SSH setup solves and what it leaves open

OpenSSH keys and Linux users solve the login problem well. They check who can log in to a host and which OS account they use. Certificates also stop old keys from working forever.

Still, the native setup leaves a major gap during the live session. Once the SSH session opens, the host trusts every command sent over that link. Standard Linux permissions apply to the whole account, not to single agent prompts. A user might allow the agent to check disk space on a prod server. If the agent makes a mistake, it might also delete files in that same folder.

Also, the remote host returns raw text to the editor. If a log file holds emails, phone numbers, card numbers, or national IDs, the agent gets them. The remote daemon does not rewrite those matched fields in the command output. Teams also lack a clear log that ties each remote command to a user identity.

Inspect one-shot SSH commands and file tasks

SSH certificates and host rules decide who is allowed on a machine. Setup alone does not check single commands or block sftp reads on sensitive paths. By design, hoop.dev runs an inline open-source SSH listener on the host where engineers and AI agents keep their normal ssh, sftp, and scp tools. On every production command or file operation, hoop.dev decides whether to pass, mask, block, or hold it for approval.

First, hoop.dev guardrails check one-shot commands and file paths before they run, and hoop.dev data masking masks matched values in file downloads. The audit trail logs the certificate principal, requested action, and verdict without storing session content. Interactive shells carry no guardrails and can be turned off when every action must be checked. Explore the code on GitHub.

Frequently asked questions

Did Windsurf change its name to Devin Desktop?

Yes, Cognition renamed Windsurf to Devin Desktop on June 2, 2026. The editor keeps the same interface and features, and old doc links redirect to the new site.

Can the editor run SSH commands without an MCP server?

Yes, the built-in agent can run terminal commands directly with user permission. Its dedicated shell loads your local profile, so it can use your local SSH client and keys.

Why are static SSH keys risky for AI coding agents?

Static keys sit on the laptop disk and rarely expire. If an agent or script misuses the key, it gets full access to the remote Linux account.

Save the open-source gateway.

hoop is MIT-licensed infrastructure for controlling how AI agents reach production data. Star hoophq/hoop so you can inspect it, deploy it, or share it when your team starts governing agent access.