mobius

How it worksPricingDocsAboutRequest access
Request access
Pricing

Pay for what your agents do, and only that.

Mobius is free for the first 90 days of early access. After that, paid by default; early-access accounts are grandfathered into discounted terms documented per customer.

Promise

The Mobius Promise

Five commitments Mobius makes in writing. The cost of each is the point: they limit what we can charge for, by design.

  • We never meter execution on infrastructure you pay for.

    Workers run where you put them. If you operate the hardware, we don't bill its runtime. By policy, not a rate card.

  • Workers route LLM traffic; we don't see your model API calls.

    Customer-managed model calls (Anthropic key, OpenAI key, self-hosted Llama via Ollama) execute on your worker. Mobius doesn't proxy them.

  • No auto-merge, auto-send, or auto-take without an approval step you wrote.

    The agent doesn't press the red button. request_approval, request_review, and request_information are step types. The run pauses, you decide, the run resumes.

  • Cost caps are first-class. Per run, per agent, per project, per day.

    Visible before the bill arrives. The ceiling is shown at the top of every run, even when nowhere near it.

  • Founder-led access for the first cohort. You'll talk to Curtis.

    Every early-access request gets a real reply from a real person. No bot, no marketing-qualified intake.

Pricing shape

Per-seat plus per-managed-run, with cost ceilings.

Pricing finalizes at GA. Plan around the shape below; your early-access account stays on early-access terms until you opt into a paid plan.

Per seat

Engineers who configure agents, automations, and workers, and the humans who approve interactions.

Per managed run

Runs Mobius coordinates. Suspended-while-waiting time doesn't bill on hold-open compute.

Cost ceiling per run / agent / project / day

Declared up front. Surfaces in the event stream as it's approached.

No metering on customer-owned infra. No credits. No opaque tiers.

Early access · free for 90 days

What you get during early access.

Full access to the platform while we firm up the commercial model. We'll let you know the production plan well before anything changes for your account.

  • Durable runs that survive restarts and resume from waits.
  • Bounded agent steps with tools, alongside deterministic action steps.
  • Sleeps, event waits, and interactions as first-class step types.
  • Workers that execute actions and customer-managed model calls in your network.
  • Environments per run, per agent, or by automation. Sprites.dev is the default provider.
  • Artifacts produced by agents (markdown, images, audio, video) saved and inspectable.
  • HTTP and SSE event streams for every run.
Request access
Talk to Curtis at curtis@deepnoodle.ai for self-hosted or enterprise.

How the bill will work (in spirit)

A base subscription plus usage on the axes above. Hard numbers will be published before early access converts to paid; your account moves to a paid plan only when you opt in.

Enterprise & self-hosted

For teams with regulatory or data-residency constraints, Mobius can be deployed in your environment. Reach out; this is a managed conversation during early access.

FAQ

Pricing and access, briefly answered.

How do I get access?

Send Curtis a note about what you'd put on a schedule, an event, or inside your product. Early access is a managed conversation; he'll get back to you directly.

Will you ever charge for execution on our own infrastructure?

No. If your worker runs on hardware you operate, we don't bill its execution time. The Mobius Promise on this page is the written commitment, dated and signed by the founder.

What does "grandfathered into discounted terms" mean concretely?

When early access converts to paid (after the 90-day free window), your account stays on the early-access discount we agreed in writing during onboarding, not the standard GA rate card. Discounts are negotiated per customer and stay with the account.

Where does my agent code run?

Agent loops and the public surface run in Mobius Cloud. Workers run wherever you put them: laptop, container, Cloud Run, on-prem. Workers connect outbound; no inbound exposure required.

Where do my secrets, tools, and customer data live?

You decide what crosses the network boundary. Sensitive code, data, and self-hosted models stay where your worker runs. Mobius dispatches each action to your worker; the worker executes it locally and returns the result. Mobius stores run state, agent and action results, and artifacts in the cloud.

Do conversations count as runs?

No. A 1:1 conversation with an agent is durable chat, separate from an automation run. If a conversation produces work that needs tracking, the agent or user creates a run, interaction, or artifact explicitly.

What happens when a run is waiting?

A run that suspends on a sleep, an event, or an interaction stops burning active work. It resumes when the condition resolves or the timeout fires.

Can I bring my own model?

Yes. Customer-managed model calls (for example, an Ollama deployment) are routed through your worker.

Is self-hosted Mobius available?

For teams with regulatory or data-residency constraints we can talk about a managed self-hosted deployment. Reach out; this is a sales conversation during early access.

Free, 3-page PDF

Get the Agent Reliability Checklist

Grade your current setup (DIY script, Routine, Codex/Cursor Automation, n8n hybrid) against the eight things that decide whether an unattended agent survives the night.