For the developer who is also the ops team

Everything around your code, at a glance.

Machines, CI, containers, spend, vendor status and agents in one window you leave open on a second monitor. Free and open source for macOS and Windows. No account, no sign-up, no server.

Latest release · macOS 14+ universal · Windows 10/11 x64

The Solador overview. Two machines connected, five GitHub repositories with one failed build flagged under Needs Attention, three idle runners, every vendor operational, and no scheduled job in trouble.
One build failed. It is the only red on the screen.
The Solador mark: a 3×3 grid of tiles in terracotta, cobalt and amber, the top-right tile larger and rotated.
The name

One tile out of true

The mark is a tiler's 3×3 grid with one tile out of true, and the name is Spanish: solar, to floor, to pave, to tile, plus -dor, the tradesperson who lays them.

The odd tile is the whole point. A grid where everything is square says nothing; what you want from a wall of panels is to notice the one thing that isn't right. So Solador stays quiet while things are fine, and the one thing that isn't is what you see.

What it shows

Eleven panels, one glance.

It opens on compact tiles you can move, hide, duplicate and scope. Open any tile for its detailed panel, in the tile's Full view.

  • A host card: overall and per-core CPU sparklines.

    Hosts

    CPU, memory, disk, network, GPU and battery for this machine, plus any host you can reach that runs the agent.

  • Containers and VMs on this machine and on a remote host.

    Containers / VMs

    docker, podman and tart, locally and on every host.

  • GitHub repositories in a section per owner, acme and globex, with workflow health and counts; one repo's build failed.

    GitHub Repos

    Workflow health for every repo you watch, grouped by the organization or account that owns it, with its open issues, ready work and pull requests (Dependabot's counted apart) beside it, and branches and worktrees in the full table.

  • Self-hosted runners grouped by OS, all idle.

    GitHub Runners

    Your self-hosted runners, with an absence roster for the ones that should be there and aren't.

  • A Claude plan with a meter for each weekly limit, the month's API spend, and this machine's Claude Code token counts.

    Anthropic

    Set up what you use: your Claude plan's limits, with whether this pace runs out before the reset; your organization's API spend; this machine's Claude Code tokens, from the logs on your disk.

  • A Codex plan with a meter for each limit, and the month's API spend.

    OpenAI

    Your Codex plan's limits and your organization's API spend, set up the same way.

  • Neon compute, storage and charges, Sentry's error count against its quota, and Vercel's month-to-date spend by service.

    Usage

    Neon, Sentry and Vercel consumption, each from its own optional credential: compute, storage and charges, errors against your monthly quota, and month-to-date spend by service.

  • Azure month-to-date cost against a budget.

    Azure Cost

    The daily cost export, month to date, read with the az CLI you are already signed in to.

  • Sentry cron monitors, all healthy.

    Sentry Crons

    Every cron monitor that is not ok, and how long it has been broken.

  • Vendor status: GitHub, Anthropic, Vercel, Neon and Azure, all operational.

    Service Health

    Status pages for the vendors you already use, with a desktop notification when one changes.

  • An OpenClaw agent farm: agents, cron and channels.

    OpenClaw

    An OpenClaw agent farm over a live WebSocket, the one panel that is event-driven rather than polled.

The first minute

A fresh install reads nothing until you add it.

Open Add source and pick what you want. These need no token and no account. An upgrade keeps what it already watched.

This machine

Per-core CPU, memory, disk and network rates, GPU and battery, sparklined at 1 Hz.

Your containers

Whatever docker, podman or tart you already run.

Vendor status

The status pages of the vendors your cockpit already reads, derived from the sources you add rather than shipped as a list.

Claude Code usage

Add Anthropic and pick Local usage: token rollups folded up from the logs already on your disk. No key, and nothing leaves the machine.

Everything after that is opt-in, one credential at a time. A panel you haven't set up shows a muted line saying what it wants, never a red error. Setup is not failure.

The one design rule

Unknown is never rendered as zero.

Every value Solador might fail to measure stays optional the whole way down, so a reading nobody could take never shows up as a reassuring zero. A cockpit you glance at has to be one you can believe.

  • —Nobody could find out.
  • 0A dimmed zero: there are genuinely none.
  • ≈Inferred, not observed, and tinted amber to say so.
  • emptyAn empty green panel is treated as a bug: it looks exactly like one that has never read anything.
Your data stays yours

No telemetry, no analytics, ever.

Apart from the vendors you configure, with your own credentials, the app talks to nothing.

  • Credentials in your OS keychain

    Keychain on macOS, Credential Manager on Windows, never the settings file. GitHub access is a read-only, fine-grained token.

  • Crash reports only if you ask

    Off until you turn it on. Even then, a report is rebuilt from an allow-list of named fields: no host names, tokens or absolute paths.

  • Signed, end to end

    A notarized macOS app and a signed Windows installer. Agent binaries are signature-checked before they run, and a paired host is reached over TLS pinned to the certificate you approved.

Get it

Give your second monitor something to say.

It ships, it's signed, and it runs all day, and it is still young. If it doesn't fit your stack, say where: that is the feedback that shapes it.

Latest release · Source on GitHub