moulds.ai
How it works

Spawn. Review. Run. Approve.

The whole flow in about 65 seconds — every clip below is a screen recording of the production product, sped up where the caption says so.

  1. 1

    Spawn — type the goal in plain English

    You describe the job the way you'd brief a person. The composer picks specialist agents from the catalog, wires them into steps, and sets the schedule — watch it happen below.

    Typing a plain-English goal into the Spawn a worker box on moulds.ai, then watching the composer work through its stages until a draft worker is ready to review.
    Sped up — real run: 3m 53s · recorded Jul 2026
  2. 2

    Review the draft — nothing runs until you say so

    The spawn ends with a draft, not an action: the worker's steps, cadence, and approval gates laid out for you to read. You edit or approve — the worker only exists once you accept it.

  3. 3

    Hire a specialist — and give it the file it works from

    Some jobs come with a specialist already built. You hire it from the catalog, set the cadence, and hand it the file it reads — here, a week's aged-receivables export. It works out what changed since last time before it drafts a word.

    Hiring Collections Clerk from the moulds.ai catalog, then dropping this week's aged-receivables export into its workspace: it works out which invoices were paid, what is still outstanding, and how many reminders are now due.
    Sped up — real run: 0m 58s · recorded Jul 2026
  4. 4

    Run — and approve what matters

    On a run, the step list advances in plain English. The run parks at the approval step the platform enforces in front of the send, and waits for your explicit approve — no approve, no send.

    The weekly Collections Clerk worker running on moulds.ai: it reconciles this week’s export, drafts the reminder each overdue invoice is due, then stops and waits for the owner — and only after Approve & continue does anything send.
    Sped up — real run: 2m 12s · recorded Jul 2026
  5. 5

    The result — with receipts

    Every finished run keeps its full trail: each step's output, and every single AI call with model, tokens, latency, and the exact prompt and response.

    Checking a finished moulds.ai worker run: every step in plain English, then one step’s technical drawer opened on its AI-calls table — the model it used, tokens in and out, retries and latency — with the exact prompt it sent and the exact response it got back expanded underneath.
    Sped up — real run: 0m 21s · recorded Jul 2026
Answers

Frequently asked questions

Every answer below describes how the product behaves today.

What does it cost to start?

Nothing — every new account starts with 200 free credits, no card required. One credit is worth $0.01. When you need more, credit packs start at $5, or you can bring your own AI keys and skip platform credits for AI calls entirely.

What if a worker makes a mistake?

Any step that sends something outward gets an approval step in front of it — enforced by the platform, not by convention: a worker without the gate can't be saved, and the engine refuses to run an ungated send. That covers workers composed for you, hand-written ones, installed ones, and scheduled runs. The run parks there until you explicitly approve it. And every run keeps a full audit trail: a plain-English step list plus every AI call with its exact prompt and response, so you can see precisely what happened and why.

Do I need to know how to code?

No. You describe the job in plain English; the composer drafts the worker — steps, schedule, approval gates — and you review the draft before anything runs. You can also install a published worker from the catalog and run it under your own account.

Can I use my own AI keys?

Yes. Connect your own keys (bring-your-own-key) and AI calls run at no platform credit cost. Without your own keys, platform-provided keys are used and metered against your credit balance, and a run stops with a clear credits-exhausted error if the balance runs out.

Ready to spawn your first worker?

Start free