Six months in: 19 runtimes, a desktop app, an Agent Builder, and a device.

Six months ago, ClawMetry was small. A dashboard that watched OpenClaw sessions. One runtime. One window. A toddler of a product, held up by the people who cheered on launch day and stuck around after.

We finished #5 on Product Hunt that week. Six hundred thousand installs later, teams inside serious companies are on trial. Some of them are the ones building the AI agents you use every day.

A lot changed to get here. This is the short version, in the order it happened.

The problem was never about one runtime.

Every team we talked to was running two, three, sometimes six different agent runtimes. Claude Code for the engineers. Cursor and Codex for the day-to-day. Something homegrown on OpenClaw or n8n behind the scenes. GitHub Copilot for the ones still writing keyboards.

And every single one was invisible in its own way. Different cost tabs. Different logs. Different mental models. Nobody could answer the simple question: what did our agents do today, and what did they cost.

So we made ClawMetry watch all of them.

Nineteen runtimes. One dashboard. One cost math. One approvals queue.

Three things we shipped in the last six months.

We did not want to just ship a bigger dashboard. Observability is a means, not the point. The point is you should be able to run more agents, with less worry, from wherever you happen to be sitting. So we opened three doors.

01. The desktop is here.

For six months we shipped a command. Fine for engineers. Not fine for the person next to them. So we made the app. Double-click, it opens. Every agent on your machine, across every runtime, already inside. It updates itself, quietly, in the background. You never think about versions again.

02. Describe the agent. We build it, host it, watch every move.

Every AI agent product you have built in the last year probably started the same way. A script, a prompt, a scheduler, a lot of glue. Then a rewrite. Then another rewrite.

Agent Builder is the last version of that story. You describe what the agent should do. We pick the right engine. We host it. We connect it to your channels. And it reports back through the same observability you already trust for the rest of your fleet.

Free to explore, with a free first agent-month for new accounts. Full plan details on our pricing page.

03. And one more thing. We built a device.

You should not have to open a browser to know your agents are okay. You should look up, and know. That is the whole idea behind the ClawMetry Desk.

It is small. It is quiet. It sits next to your monitor and shows what matters right now. Cost today. Live agents. What is idle, what is running, what is covered by your plan, and what is drifting into your API bill. When something needs a decision, it lights up. When it does not, it is just there.

The first units are on desks with our earliest supporters. Availability is limited and staged while we harden firmware and audit control actions. If you want on the list, hit reply to the announcement email or the device page.

It stopped being a solo tool.

Every team we shipped to said the same thing. Great, but who owns this at my company, and how do I stop the AI bill from blowing up. So we built the layer above the individual dashboard.

Invite your team. See everyone's agents on one screen.

One fleet view. Every teammate's sessions, every runtime, every cost. Roles for who can approve what. SSO if you need it. Budget caps per team, per runtime, per project. Set a $2,000 monthly limit and get a ping in Slack at 50%, 80%, and 100%.

Nobody discovers the AI overrun in next month's invoice anymore.

Govern agents with a control plane.

Before an agent runs a risky shell command, places a physical order, or sends an email to your investor, someone gets a chance to say yes. Approvals route where your team already lives. PagerDuty for the ones that page. Slack for the ones that ask. Webhooks for everything in between.

PagerDuty Slack Webhooks SSO / SAML RBAC Audit logs

The individual dashboard is still free forever. The team layer is where Pro earns its keep.

Six months in numbers.

None of these numbers are ours. They belong to the people who installed the thing, filed the bugs, and shipped the fixes.

Enterprise teams are on trial now. That is a sentence that would have made no sense on Feb 18.

For the people reading this today.

We would not be here without the people who kept installing, kept telling us what to fix, and kept forgiving the beta rough edges. So this post has thank-yous baked in.

If you got the email, hit reply. If you did not and want in, email us here and tell us which of the three (desktop, Agent Builder, Desk) you would use first, and what you would want it to do that it does not do yet.

Thanks for being here through the messy middle. The next year is going to be the fun part.

Vivek
Founder, ClawMetry