Get early accessEarly access
Free during the preview

Your AI dashes to every machine.

Dashcat lets your AI check, fix and update the Windows, Mac and Linux machines you look after, from your laptop, browser or phone. Install Dashcat on each machine, connect your AI once, then ask in plain words.

Get early accessSee what it can do
  • Windows · macOS · Linux
  • About 5 MB idle
  • Works with any MCP AI
Your AIconnected through Dashcat · 5 machines
5 of 6 machines onlineactions today 128
Your AIvia Dashcat
Activity logevery action, on every machine

    Works with the AI you already use

    • ClaudeWeb, desktop, mobile and Claude Code
    • ChatGPTCustom connectors
    • CursorMCP servers
    • VS CodeCopilot agent mode
    • CodexCLI and IDE
    • Gemini CLIMCP servers
    • Local agentsAny agent that speaks MCP

    Dashcat speaks MCP, the open standard for connecting AI to tools, so any AI that supports it can use your machines.

    Support every machine without visiting one.

    Anything you would do sitting at that computer, your AI can do through Dashcat: find out what is wrong, change settings, install software, run programs and report back.

    “Why is the reception PC so slow? Fix what you can.”

    Your AI checks what starts at login, what is filling the disk and what needs updating, sorts it out and tells you what it changed.

    Works today on your network

    “Which laptops are missing this month’s updates?”

    One question, asked of every machine, with one clear answer back instead of a dozen remote sessions.

    Works today on your network

    “Install the new printer driver on every PC in accounts.”

    Carried out on each machine, with the result from every one of them and a record kept on each.

    Works today on your network

    More use cases: small business, software teams, home and family, home labs

    Set up once. Ask from anywhere.

    How it works
    1. 1

      Install Dashcat on each machine

      A small program that sits in the tray and uses about 5 MB of memory. Windows, macOS and Linux, Raspberry Pi included.

    2. 2

      Connect your AI

      Add the Dashcat connector to your AI once, and choose which machines it may reach. Nothing to install where your AI runs.

    3. 3

      Ask in plain words

      From your desk, your browser or your phone. Name the machine, or several, and say what you need.

    Your AI does the work. You stay in charge.

    Your AI reaches only the machines you connect, and only while Dashcat is running on them.

    You choose the machines

    A machine ignores everyone it has not been connected to. You decide which machines each AI may reach, and take access back with one click.

    Every action on the record

    Each machine keeps a log of what was run on it, and by whom. Read it on the machine, or all in one place.

    Direct mode, when it must stay private

    For sensitive machines, Dashcat runs end to end from your own computer instead, and the cloud cannot see or run anything. Mutual TLS 1.3, keys pinned at pairing.

    How Dashcat keeps your machines safe

    For developers

    An MCP server for every machine you own.

    Dashcat gives Claude Code, Codex, Cursor and any MCP client tools to run commands, manage long-running programs, move files, follow logs, take screenshots and drive desktop apps on your nodes.

    # Direct mode, from your terminal
$ claude mcp add dashcat -- dashcat-bridge mcp
    • Mutual TLS 1.3
    • Keys pinned at pairing
    • SPAKE2 pairing
    • About 5 MB idle, one thread

    Read the quickstartEvery tool and its limits

    claude · ~/src/app on mac-studiowin-build connected

    > The Windows build is failing on win-build. Fix it.

    ● dashcat.run win-build cargo build --release
    exit 101 error[E0433]: could not find `unix` in `os`
      --> src\perm.rs:3:14

    perm.rs uses a Unix-only API with no Windows branch. Gating it behind cfg(unix) and writing the change to the node.

    ● dashcat.write_file win-build C:\src\app\src\perm.rs
    wrote 1.2 KiB (overwrite)

    ● dashcat.run win-build cargo build --release
    exit 0 Finished `release` profile in 48.2s

    Fixed. The build passes on win-build, and the same change is in your working tree here.

    An example session. Your AI client names the node; Dashcat runs the call there.

    What your AI can do on a node

    Commands

    run · node_info

    Up to 30 minutes per command, with output capped at 64 KiB per stream.

    Programs

    start_process · process_output · stop_process

    Dev servers and watchers keep running. Stopping one stops everything it started.

    Files and logs

    read_file · write_file · tail_file

    Follow a log as it grows, and push builds to nodes with dashcat-bridge push.

    Screen and UI

    screenshot · ui_tree · ui_action

    See the desktop and press real buttons through each OS's accessibility interface.