Ssh MCP server
{
"mcpServers": {
"pulsemcp-ssh": {
"command": "npx",
"args": [
"-y",
"ssh-agent-mcp-server@latest"
],
"env": {
"AGENT_ORCHESTRATOR_BASE_URL": "https://...",
"FLY_IO_API_TOKEN": "<fly_io_api_token>"
}
}
}
}{
"mcpServers": {
"pulsemcp-ssh": {
"command": "npx",
"args": [
"-y",
"ssh-agent-mcp-server@latest"
],
"env": {
"AGENT_ORCHESTRATOR_BASE_URL": "https://...",
"FLY_IO_API_TOKEN": "<fly_io_api_token>"
}
}
}
}{
"mcpServers": {
"pulsemcp-ssh": {
"command": "npx",
"args": [
"-y",
"ssh-agent-mcp-server@latest"
],
"env": {
"AGENT_ORCHESTRATOR_BASE_URL": "https://...",
"FLY_IO_API_TOKEN": "<fly_io_api_token>"
}
}
}
}{
"servers": {
"pulsemcp-ssh": {
"command": "npx",
"args": [
"-y",
"ssh-agent-mcp-server@latest"
],
"env": {
"AGENT_ORCHESTRATOR_BASE_URL": "${input:agent_orchestrator_base_url}",
"FLY_IO_API_TOKEN": "${input:fly_io_api_token}"
}
}
},
"inputs": [
{
"id": "agent_orchestrator_base_url",
"type": "promptString",
"description": "Agent orchestrator base url",
"password": true
},
{
"id": "fly_io_api_token",
"type": "promptString",
"description": "Fly io api token",
"password": true
}
]
}{
"mcpServers": {
"pulsemcp-ssh": {
"command": "npx",
"args": [
"-y",
"ssh-agent-mcp-server@latest"
],
"env": {
"AGENT_ORCHESTRATOR_BASE_URL": "https://...",
"FLY_IO_API_TOKEN": "<fly_io_api_token>"
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y ssh-agent-mcp-serverversion 0.1.5
What it does
This MCP server provides tools for managing SSH remote servers with SSH agent authentication support. It is for developers who need to execute commands and manage files on remote systems through an LLM-powered interface.
- Execute commands on remote SSH servers and retrieve their output.
- Upload or download files between local and remote SSH servers.
- Manage SSH connections and authentication using SSH agent credentials.
This entry comes from the registry but points to no public repository, so there is nothing to check: no commit history, no source, no tool list. It is listed for completeness and left out of search engines.
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.