What it does
A single dashboard panel that shows every machine connected as a Multica runtime — local laptops, cloud VMs, anything running the daemon. Real-time monitoring (online/offline status, usage charts, weekly activity heatmap), per-runtime token spend (input / output / cache read / cache write), auto-detection of installed agent CLIs.
What a runtime is (per the docs)
"A Runtime is a compute environment that can execute agent tasks. It can be your local machine (via the daemon) or a cloud instance. Each runtime reports which agent CLIs are available, so Multica knows where to route work."
So a runtime is a labeled execution target the orchestrator can dispatch tasks to. The agent itself doesn't choose where it runs; the workspace assigns it to a runtime.
Captured UI elements (from landing screenshot)
- Runtime card titled
MacBook Pro·online·arm64 / macOS 15.2 - Token usage breakdown:
Input 2.2M · Output 1.1M · Cache Read 1.5M · Cache Write 338.0K - Weekly activity heatmap (Mon-Fri × hour grid)
- Daily cost chart spanning Mar 18 → Mar 31
Target ICP
Both ring 1 (solo dev with a laptop) and ring 2 (small team with multiple connected machines) and ring 3 (self-hosted with cloud runtimes).
Differentiators
- Local + cloud in one view — no context-switch between Modal-style cloud dashboards and local tooling.
- Auto-detection — runtime reports which agent CLIs are available (Claude Code, Codex, Cursor Agent, OpenCode, OpenClaw, Hermes, Gemini, Pi, Kimi, Kiro CLI). "Connect a machine, and it's ready to work."
- Token-aware cost surface — per-runtime spend is the primary metric, not request count. Targeted at the buyer who already feels Anthropic / OpenAI bills.
Strategic implications for CreateOS
CreateOS could ship a Multica-compatible cloud runtime adapter and instantly become a deployment target inside Multica's runtime panel. The integration shape: a CreateOS sandbox advertises itself to a Multica workspace as a runtime; tasks dispatched by the assignee picker land in the sandbox; MPP earnings on agent work flow back as a metric on the runtime card. This is a real BD pitch — Multica's daemon model is designed for this kind of pluggable runtime.
Closest analog: how Vercel Sandbox / Modal advertise themselves to the OpenAI Agents SDK as default providers. Same shape, different listing.
Sources
- multica.ai homepage (Runtimes section + screenshot)
- multica.ai/docs (Welcome — runtime model definition)