Browser Ssh

What the browser-ssh MCP server does

The browser-ssh MCP server gives an AI agent access to saved SSH servers through three stated capabilities: executing SSH commands, working with SSH sessions, and transferring files through SFTP. Access is described as scoped and audited, so the server is intended for controlled remote operations rather than unrestricted credential sharing.

The saved-server model is central to the described workflow. Instead of requiring an agent to receive or handle connection credentials directly, the server works with server entries that have already been saved. The supplied description does not specify how those entries are created, what scope controls are available, or how audit records are presented.

This fits use cases where an agent needs to inspect or operate a remote host, maintain an SSH session, or move files over SFTP. It may also suit environments where credentials must remain outside the agent’s direct context.

How it works

An agent accesses the browser-ssh MCP server through MCP and requests one of the documented remote-operation categories: SSH execution, SSH sessions, or SFTP. The server applies the stated scope around access and records activity for auditing. The connection target comes from the set of saved servers rather than from credentials supplied in the agent conversation.

The available material does not describe the MCP transport, request schemas, authentication model, session lifecycle, or audit-log format. It also does not explain whether command execution is interactive, how file paths are handled, or whether multiple saved servers can be selected in one workflow. Those details should be confirmed before implementation.

Tools and capabilities

• SSH command execution on saved servers.
• SSH sessions for work that requires a session rather than a single command.
• SFTP operations for remote file transfer.
• Scoped access intended to limit remote actions.
• Audited activity for remote operations.
• Credential handling that does not expose server credentials to the agent.

Limitations and notes

The provided material identifies capabilities but does not include installation instructions, configuration names, client compatibility, licensing, permission syntax, or operational limits. No claims can therefore be made about support for a particular editor or MCP client, required environment variables, deployment modes, or authentication providers.

Before adopting the browser-ssh MCP server, confirm how saved servers are provisioned, how scopes are defined, where audit data is stored, and which SFTP actions are available. Also establish whether SSH sessions persist between tool calls and how failures, timeouts, and host-key verification are handled. These questions matter when the server will be used for production administration or access to sensitive systems.

The browser-ssh MCP server should be evaluated against the required control boundary: the supplied description promises scoped and audited operations without exposing credentials, but it does not define the exact enforcement mechanism or provide evidence about isolation. Treat those implementation details as items for project-level verification.