Setup

From first launch to a full cockpit

Solador is useful the moment it opens. Everything after that is opt-in, one credential at a time, in any order. A panel you haven't set up says what it wants in a muted line, never a red error.

  1. Install it and open it

    Download the signed installer for macOS or Windows. A fresh install reads nothing until you add it, and an upgrade keeps what it already watched. Under Settings → Sources → Add source, these need no token:

    • This machine: per-core CPU, memory, disk and network, GPU and battery, and your local docker, podman or tart containers;
    • Anthropic, with Local usage picked: your Claude Code token usage, read from the logs on your disk;
    • Service health: the status pages of the vendors you add.
  2. Connect GitHub optional

    For the Repos and Runners panels, create a fine-grained personal access token. Solador never writes to GitHub, so everything is read-only:

    • Repository permissions: Actions, Contents, Issues and Pull requests (Metadata comes with them).
    • Organization permissions: Self-hosted runners, for the Runners panel, and Projects, for the Repos panel's Ready column.
    • Everything else stays at No access.

    Then in Solador: Settings → Sources → Add source → GitHub. Name the account and paste the token into Fine-grained PAT, list the organizations whose runners you want on the next step, and choose Finish: nothing is saved before that. On the account's card, Add a repo finds the repositories to watch as you type.

    The full walkthrough, with what each permission feeds: docs/github-setup.md.

  3. Add your other machines optional

    Another machine (a Linux server, a second Mac) shows up in your cockpit by running the agent. It runs as your user, with no sudo, on Linux with systemd or macOS 11 and later. It needs minisign to check the release's signature: brew install minisign, apt install minisign (Debian 12+, Ubuntu 24.04+) or dnf install minisign.

    On the host:

    curl -fsSLo bootstrap.sh https://raw.githubusercontent.com/Sassy-Dog/solador/main/agent/deploy/bootstrap.sh
    bash bootstrap.sh

    It fetches the installer from main, verifies the agent binary's signature before anything runs, installs it, and asks for a bearer token (press Enter to generate one). It finishes by printing the agent's certificate fingerprint. Two steps rather than a curl | sh, deliberately, so you can read what you run.

    Then in Solador: Settings → Sources → Add source → Remote host. Enter a name, the host's address and the token, then Check certificate. Compare the fingerprint it shows with the one the installer printed, choose Trust, and Add Host.

    Or let a coding agent do it

    The app is also a command-line tool that adds a host over your own ssh. It reads the agent's fingerprint and token on the host and hands both to the open app, which checks the certificate and the token itself before it saves anything. It changes the host only where a flag allows it, and never runs sudo. Give this to a coding agent (Claude Code, Codex, …) with your host filled in. The prompt is written for macOS. On Windows the same command runs from solador-app.exe in Solador's install folder: put its full path in the prompt's second line (and "this PC" for "this Mac" in its first). It uses Windows' own ssh.exe, which has not been tried on a real Windows machine yet.

    Add <user>@<host> to Solador on this Mac with Solador's own command-line tool.
    Below, solador-app means /Applications/Solador.app/Contents/MacOS/solador-app;
    if Solador is installed somewhere else, ask me where.
    
    1. Run: solador-app host add --ssh <user>@<host> --dry-run
       If it refuses and its message says to pass --install, --enable-tls,
       --bind-all or --address, show me the message word for word and ask me
       which, if any, to add: --bind-all has the agent listen on every network
       the host is on, a public one included, with only its token guarding it.
       --address takes a value: use the one the message names, or ask me; never
       guess one. Add only what I agree to, then run the dry run again. Repeat
       until it prints "Would pair ...". If a refusal names a flag the command
       already has, show me the message and stop.
    2. Show me that plan, every "first:" and "note:" line included, and wait for
       my go-ahead. Then run the same command, with the same flags, without
       --dry-run.
    3. Finish with: solador-app host list
    
    Rules:
    - Solador must be open. If the tool says it is not, ask me to open it. Never
      edit Solador's files or its keychain item.
    - Never print, log or ask me for the agent's bearer token.
    - Never use sudo. If the tool exits 3, show me the step it printed, wait for
      me to run it, then run the same command again.
    - Any other refusal (an unknown ssh host key, an agent run by another user,
      SOLADOR_AGENT_REQUIRE_TAILNET, a Mac user with no login session): show me
      its message and stop. Re-run as another ssh user only if I say I can log
      in as them.
    - Never remove SOLADOR_AGENT_REQUIRE_TAILNET, and never rotate or regenerate
      a token. If any message mentions --replace, show me and stop: pass
      --replace only if I ask.
    - Do not accept a new ssh host key, edit ssh config, or install packages on
      the host.

    The command reference has every flag.

    Any network path works: a LAN, a VPN, Tailscale, an SSH tunnel. A paired host is reached over TLS pinned to the certificate you approved, and Solador never sends the token over plain HTTP anywhere but loopback or Tailscale. A fresh install serves TLS, and on a machine without Tailscale it listens on all interfaces, behind TLS and the token: firewall its port on a machine with a public address. On a machine with Tailscale it listens only on its Tailscale address, so a computer that is not on your tailnet sees “not paired” for its LAN address until you bind all interfaces and pair that address. The agent reference covers binding, updates, unattended updates and uninstalling.

  4. Connect the rest optional

    Each is independent, under Settings → Sources → Add source:

    • Neon: an organization API key and your organization ID, for compute, storage and estimated charges. Nothing is saved until you press Finish.
    • Sentry: an org:read token and your org slug. One credential feeds both Usage and Sentry Crons, and the steps ask which of the two to turn on. Nothing is saved until you press Finish.
    • Vercel: a token and, for a team, its team ID, for month-to-date spend from the billing export. Nothing is saved until you press Finish.
    • Anthropic and OpenAI: each walks you through what you want. API key needs an Admin API key, which reads your organization's spend this month. Anthropic's starts sk-ant-admin and only a Console organization can create one, and an ordinary API key is refused.
    • Claude and Codex plan limits: nothing to paste. Pick Subscription in those same steps, and that tile reads the tool's own saved sign-in for the plan's limits: a meter for each 5-hour and weekly window the plan has (or an Enterprise seat's budget), with whether the current pace runs out before the reset. For Claude it also reads the plan's name from ~/.claude.json, keeping that one field and nothing else. It only reads: the sign-in is never copied, stored, refreshed or changed. On a Mac, macOS may ask whether Solador may read Claude Code's keychain item. Both plan endpoints are undocumented and may change.
    • Azure Cost: the storage account and container of a daily cost export, and optionally a monthly budget. Nothing to paste: each read is signed by the az CLI you are already signed in to. Nothing is saved until you press Finish.
    • OpenClaw: your gateway's URL. Solador shows the command that approves this device on the gateway, then follows it over a live WebSocket.
    • Status page: any Atlassian Statuspage you depend on, watched beside the vendors your other sources add.

    Credentials go to your OS credential store (Keychain on macOS, Credential Manager on Windows), never the settings file.

  5. Make it yours

    Edit dashboard moves, hides, duplicates and scopes the overview's tiles. Each tile's Configure has a Full view page for how its panel is laid out. Then put it on your second monitor and leave it open.

    Doesn't fit your stack? Start a discussion or open an issue.