Only what worksis listed as working.

Each integration below runs today and says where it is checked. The planned list is separate and stays planned until it runs.

Works today

01Contract calls on Robinhood ChainA rule per contract and function, with an ETH cap per call. The rail checks it on every run; a manifest update voids old rules.LauniqTaskRail
02Fixed arguments, pinned on chainThe studio encodes them and the task stores their hash. A run with any other call data is refused (code 7).Studio + rail
03A number from an HTTPS JSON sourceThe runner fetches it when the task runs (https only, 8 s, 64 KB), reads one path, scales it to whole units, and builds the call. The rail still checks the call.Runner (your machine)
04SchedulesRuns, interval, start and deadline live in the task on chain; too early or too late is refused whatever the runner's clock says.Rail
05Owner approval of exact callsThe runner proposes call data and ETH value; the owner approves that hash; only that call can run, once.Rail + dashboard
06Your own code in the runnerTasks with free arguments call decideArgs() in the runner, which you write. The rail limits what it can do; token-moving calls wait for approval.Runner (your code)

Planned, not built

PlannedHosted runnersRunning agents for people means holding runner keys and paying for uptime. Not before a security review and a clear cost model.
PlannedNotifications (chat, email, webhooks)Today the dashboard and the runner log are the only alerts.
PlannedShared manifest libraryA place to publish and reuse manifests, with their hashes. Not built.
PlannedOther chainsRobinhood Chain only today.
PlannedModel providers inside LauniqNo AI model is bundled or called by Launiq. You can call one from your own runner code.
Planned$LAUNIQ creditsPlanned payment for the hosted parts above, once they exist. Not needed for anything today.