How to Configure JetBrains AI SSH MCP Server

Published
Length
3 min read

Luis Felipe Dias

All posts

You set up a jetbrains ai ssh mcp server by linking your local IDE assistant to an MCP process that runs remote shell commands over SSH. This setup lets the AI assistant read logs, check files, and run scripts on remote Linux hosts. Most teams start by pointing the local MCP server at a standard SSH key on the developer laptop. That quick path works at first, but broad SSH privileges let an automated agent run destructive commands across your servers. How a jetbrains ai ssh mcp server works T

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.

You set up a jetbrains ai ssh mcp server by linking your local IDE assistant to an MCP process that runs remote shell commands over SSH. This setup lets the AI assistant read logs, check files, and run scripts on remote Linux hosts. Most teams start by pointing the local MCP server at a standard SSH key on the developer laptop. That quick path works at first, but broad SSH privileges let an automated agent run destructive commands across your servers.

How a jetbrains ai ssh mcp server works

The Model Context Protocol gives IDE assistants a standard way to call external tools. In this setup, JetBrains AI sends tool requests to a local MCP server process. That local process opens an SSH session to your target Linux host and runs the requested shell command. The remote host sends the command output back through the SSH tunnel to the MCP server. Then the IDE assistant reads that output and decides its next step.

Because the MCP server uses standard OpenSSH under the hood, it relies on your local SSH setup. Teams often use raw public keys or OpenSSH certificates to log in to remote hosts. In OpenSSH, a Certification Authority key can sign a user public key to create a certificate. That certificate holds a public key, identity info, principal names, and a set of options. OpenSSH certificates use a much simpler format than X.509 certificates, and OpenSSH added support for them in version 5.4.

When an admin signs a user certificate, they set a key identity and list one or more user principals. The admin can also set a validity time window so the certificate expires on schedule. If the admin does not set a validity window, the certificate stays valid forever by default. Using short validity windows helps teams avoid leaving permanent keys on developer laptops for the MCP server to use.

What native SSH setup solves and what it misses

Native OpenSSH keys and certificates solve the host login problem well. They check who the user is before opening a shell session on the target machine. You can also route traffic through a bastion or jump host to keep private servers off the public internet. Once the SSH check passes, the remote daemon starts a shell for the mapped Linux user account.

That login check leaves a big gap when an AI agent drives the terminal. SSH decides who logs in, not what individual shell commands do in flight. If the Linux account can restart services or delete files, the AI assistant can run those commands too. A single bad prompt or misread log line can trigger a command that breaks a live system.

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.

Raw command outputs also flow straight back into the IDE context window. When an agent inspects app logs or database dumps over SSH, the output often holds sensitive data. Emails, phone numbers, card numbers, and national IDs in those logs get sent directly to the model. Native SSH has no way to rewrite matched fields in results before the MCP server reads them.

Controlling AI shell commands and sensitive output

To run an SSH MCP workflow safely, teams must separate login rights from command execution rights. First, give the AI workflow a dedicated Linux user with read-only permissions on specific log folders. Never let the MCP server use a personal developer key that has full admin rights.

Second, think about how you review high-risk actions and filter output data. Static Linux file permissions cannot pause a single shell command to ask a human for approval. They also cannot strip card numbers or emails out of a log file stream on the fly. Closing those gaps requires inspecting each command and its output between the assistant and the remote host.

Add command and file guardrails to SSH access

Host firewalls and certificate principals control who opens an SSH session. Yet an allowed user or agent can still run a bad one-shot command or pull sensitive files over sftp. In practice, hoop.dev provides an open-source SSH server with guardrails on the host that uses OpenSSH certificates.

Using hoop.dev SSH controls, guardrails check one-shot commands, env variable names, and sftp or scp paths before they run. Port forwards default to deny unless the target is listed. Because interactive shells have no statement boundary and carry no guardrails, teams can turn them off so every command is checked and logged. Check out the MIT-licensed project on GitHub.

Frequently asked questions

Does the JetBrains AI assistant connect to remote hosts directly?

No, the IDE assistant talks to a local MCP server on your machine. That local MCP server opens the SSH connection to the remote host and runs the shell commands.

Why are static SSH keys risky for MCP servers?

Static SSH keys stay valid until someone removes them from the remote server. Any local process or AI tool that can read the key file gets full access to the remote account.

Can OpenSSH certificates restrict what data an AI model reads?

No, OpenSSH certificates only control how a user or host logs in and which options apply to the session. They do not inspect command output or mask sensitive fields like emails or phone numbers in flight.

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.