
Hop
io.github.danmartuszewskiPublished by GitHub account danmartuszewskiv1.1.0Updated Oct 24, 2026
SSH connection manager with MCP server for AI-assisted operations across servers.
Overview
Hop is a local SSH connection manager whose built-in MCP server lets an AI assistant browse, search and resolve your saved server connections.
- What you can do
- List, search and inspect saved SSH connections, filtered by project, environment or tag, and view details of a single connection by ID or fuzzy match. Preview how a target pattern (named group, project-env pattern, glob or fuzzy match) resolves to connections, and list named groups. Read connection usage history and build the full SSH command string for a connection. Browse the same data as resources: config summary, all connections, individual connections and groups. Optionally, when started with remote execution enabled, run shell commands on matched servers with per-host output limits.
- When to use it
- Useful when you keep many SSH hosts in Hop and want an assistant to find the right server, check which hosts a group or pattern covers, or look up connection details without retyping SSH commands. Only worth adding if you already use Hop or plan to manage connections with it.
- Requirements
- Runs as a local process on your machine; the Hop binary must be installed (Homebrew, Go, or from source) and registered with your MCP client as hop mcp. It reads its own config file and can import hosts from your SSH config. Remote command execution requires starting the server with the allow-exec flag.
Installation
In SourceWeft
- Open Hop in the dashboard and add it to a workspace.
- Enable the server for the chats that should use its tools.
Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.
Other MCP clients
Follow the launch instructions in the repository.
README
hop
Stop typing long SSH commands. Just hop prod and you're in.
Why hop?
Install
Homebrew (macOS/Linux)
Go
From source
Install with an AI agent
Using Claude Code, Codex, Cursor, or another coding agent? Paste the block below into your agent and it will pick the right install path for your machine, register hop's MCP server, and verify the install.
Features
- Fuzzy matching - Type
hop prodto connect toapp-server-prod-03 - TUI dashboard - Browse, add, edit, delete connections with keyboard or mouse
- SSH config import - Already have servers in
~/.ssh/config? Import them in one command - Export - Export filtered connections to YAML for sharing or backup
- Multi-exec - Run commands across multiple servers at once
- Groups & tags - Organize by project, environment, or custom tags
- Jump hosts - ProxyJump support for bastion servers
- Landing directory - Drop straight into a predefined working directory on connect
- MCP server - Let AI assistants manage your servers — search connections, run commands, check status across projects
- Mosh support - Use mosh instead of SSH for roaming and unreliable connections
- Zero dependencies - Single binary, works anywhere
See all features in action: Demo recordings
Raycast Extension
Launch connections directly from Raycast. Fuzzy search, tags, environments - all at your fingertips.
Configuration
Config file location: ~/.config/hop/config.yaml
Security note:
forward_agent: trueexposes your SSH keys to anyone with root access on the remote server. Only enable this for servers you fully trust. Consider usingproxy_jumpinstead when you just need to reach internal hosts through a bastion.
Mosh Support
Mosh (mobile shell) is useful for connections over unreliable networks — it handles roaming, intermittent connectivity, and high latency gracefully.
Global default — enable mosh for all connections:
Per-connection — enable mosh for specific connections:
One-off — use the --mosh flag without changing config:
Per-connection use_mosh: false overrides the global default. SSH options (port, identity file, proxy jump, agent forwarding) are automatically passed to mosh via its --ssh flag. Mosh requires both the local mosh-client and mosh-server on the remote host.
Note:
hop execalways uses SSH regardless ofuse_mosh, since mosh is designed for interactive sessions.
Landing Directory
Set remote_dir to have a connection drop you straight into a specific directory instead of $HOME:
On connect, hop runs cd into the directory and then hands you a normal interactive login shell, so the session behaves exactly like a regular SSH login — just somewhere else. A few details worth knowing:
- Absolute paths and
~both work.~and~userare expanded by the remote shell. - Forgiving by design. If the directory is missing or inaccessible, you still land in a shell (in
$HOME) rather than getting bounced off the host. - Works in new tabs too.
hop opencarries the landing directory into every terminal it launches.
Note:
remote_diris ignored when you pass an explicit command (e.g.hop connect web -- uptimeorhop exec), since those aren't interactive sessions.
TUI Dashboard
Launch with hop or hop dashboard.
When you connect to a server from the dashboard (by pressing Enter), the SSH session starts, and the dashboard automatically returns after the session ends. This lets you quickly hop between servers without restarting the TUI each time.
For one-shot connections that exit to your terminal, use:
Keyboard Shortcuts
Filtering Connections
Press / to filter connections by typing keywords. The filter supports multi-keyword AND logic - separate keywords with spaces to find connections matching all terms.
Examples:
prod- matches connections containing "prod"prod web- matches connections containing both "production" AND "web"kaf staging- matches connections with both "kafka" AND "staging"
The filter searches across connection IDs, hosts, projects, environments, and tags.
Quick Add with Paste
Press p and paste any of these formats:
The connection form opens with fields pre-filled.
Duplicating a Connection
Press c on any connection to create a copy. The add form opens with every
field pre-filled from the original — including options that aren't shown in
the form, like proxy jump and mosh — and a collision-free ID suggestion (e.g.
web-prod → web-prod-copy). Adjust whatever you need and save. If you pick an
ID that already exists, the form stays open with your edits intact so you can
fix it.
Importing from SSH Config
Import existing connections from your ~/.ssh/config file:
From the dashboard: Press i to open the import modal, select which connections to import, and press Enter.
From the CLI:
What gets imported:
- Host alias becomes the connection ID
- HostName, User, Port, IdentityFile
- ProxyJump for jump host connections
- ForwardAgent setting
What gets skipped:
- Wildcard patterns (
Host *,Host *.example.com) - Entries without a HostName (alias is used as hostname)
Conflict handling: If a connection ID already exists, the imported connection is renamed with -imported suffix (e.g., myserver → myserver-imported).
Exporting Connections
Export a subset of connections to a YAML file for sharing, backup, or transferring to another machine.
From the dashboard: Press x to open the export modal. Only currently filtered connections are shown — apply text or tag filters first to narrow the selection. Toggle items with Space, then press Enter to save.
From the CLI:
At least one filter flag or --all is required. Filters combine with AND logic.
Theming
The dashboard ships with sixteen color presets — each popular theme has both a dark and a light variant, listed separately so you can pick whichever you want regardless of your terminal background. Press T to browse them with live preview: ↑/↓ to navigate, Enter to save the choice into your config, Esc to revert.
Picking a preset writes a single line to your config:
When theme_preset is unset, hop auto-picks default-dark or default-light based on your terminal background.
Custom overrides
Layer your own colors on top of any preset:
Color values can be either a quoted ANSI 256 code ("39") or a hex string ("#bd93f9"). ANSI codes adapt to your terminal's palette; hex values are absolute.
Available keys: primary, secondary, accent, success, warning, error, muted, selection, foreground. Any key you don't set falls through to the preset, then to the built-in default.
CLI Commands
Targeting
Commands like exec and open accept a target that resolves to one or more connections. The target is matched in this order:
- Named group — an explicit list of connection IDs defined under
groups:in config - Project-env pattern — matches connections by
projectandenvfields (e.g.myapp-prodmatches all connections withproject: myappandenv: prod) - Glob pattern — wildcard matching on connection IDs (e.g.
web*,*-prod-*) - Fuzzy match — falls back to fuzzy matching a single connection ID
You can also filter any target by tag with --tag.
Use hop resolve to preview which connections a target will match before running anything:
Examples
Scripting with hop
hop get prints connection fields to stdout so you can drop them straight into shell pipelines and command substitutions — think of it as ssh -G for your hop config.
Matching is exact ID only (not fuzzy) — safer inside scripts. Unknown IDs exit 1 with a "did you mean" hint. See hop get --help for the full field list.
MCP Server (AI Assistant Integration)
hop includes a built-in Model Context Protocol server that lets AI assistants like Claude Code and Codex manage your servers directly. Ask your assistant to check disk space across production, restart a service on staging, or find which servers belong to a project — it discovers your connections, resolves targets, and executes commands through hop.
[Claude Code managing servers through hop's MCP server]
Setup
Claude Code:
Codex CLI:
Claude Desktop — add to your config (~/Library/Application Support/Claude/claude_desktop_config.json):
Codex — add to ~/.codex/config.toml:
Or generate the Claude Desktop config automatically:
Tools
By default, only read-only tools are exposed:
To enable remote command execution, start with --allow-exec:
This adds the exec_command tool, which runs shell commands on matched servers with output limits (64 KiB each for stdout and stderr per host, 50 hosts max). Output is capped during capture; excess output is discarded and marked as truncated in the response.
Resources
The server also exposes browsable resources:
Security
- Identity files (SSH key paths) are never exposed through MCP
- Remote execution is disabled by default and requires explicit
--allow-exec - All logging goes to stderr to keep the JSON-RPC transport clean
Shell Completions
Flags
Building
Maintenance
Dependency maintenance is configured through Dependabot for Go modules (including the pinned vulnerability scanner), GitHub Actions, and Docker images. It checks weekly and opens pull requests for review. Minor and patch Go updates are grouped; major updates remain separate.
The Security workflow runs make vuln on pull requests, pushes to main, and a
weekly schedule. It can also be started manually from GitHub Actions. Reachable
vulnerabilities fail the check. Run it locally with a patched, supported Go
toolchain so the scan also reflects current standard-library fixes.
Docker
Project Structure
License
MIT License - see LICENSE for details.
Source: README.md at commit 2adeb31
Tools
0Version history
1- v1.1.0LatestSep 16, 2026
Score
50% of the applicable checks met
How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.
- Reached in the last 30 days25 pts
Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.
- Auth declared10 pts
Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.
- Tools listed15 pts
Answer tools/list without a credential, or ship a package that starts in a sandbox without one.
- DescriptionRequired15 pts
- Logo10 pts
- Documentation15 pts
- License10 pts