Today we are releasing the first version of Pulse, the Runbaze platform. The initial release is built for software delivery teams, our native ground, and it installs in the client's own environment. One installation serves one client. Your code, your data, and the record of what happened stay on your systems.
The release covers the full loop we have been writing about. Agents work tasks on a live board, and autopilot chains them: when one task completes, the next in the dependency line starts on its own. A safety rules engine ships with sixteen seed rules that block destructive commands out of the box, and takes custom rules scoped to a project or a single agent. Cost budgets warn or block before an agent spends what it shouldn't. And when an agent needs a decision, it stops and asks: a prompt appears, a person chooses, the run continues. The gate is built in, and it has a name on it.
Everything lands in the record. Every run is kept with its history, filterable by agent, status, and time. Agents save their outputs as artifacts you can open and review. A dashboard shows the leaderboard, success rate, cost, and duration. The changelog writes itself from the git log, an evaluation engine scores agents against test suites so you can tune them on evidence, and signed webhooks push run events to wherever your team watches.
Pulse connects to the systems software teams already run: Slack, GitHub, Atlassian, Linear, Notion, Sentry, Supabase, PostgreSQL, AWS, and GCP. The model underneath stays your choice.
Software teams are first because that is where AI output already outruns human review, and where the answer for it matters most. The domains on our roadmap follow the same pattern: agents on the volume, a person at the gate, a record that proves both.