moulds.ai
Now in private beta

Spawn an AI worker for the job.

Watch it work. Approve what matters.

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: 4m 1s · recorded Jul 2026
  1. 1Type the goal in plain English.
  2. 2The composer drafts a worker — steps, schedule, approval gates.
  3. 3You review the draft. Nothing runs until you say so.
  4. 4It works — pausing for your approval, every step traced.

Live counts from the public catalog — every figure links to its source.

For your job

Workers for the roles you can't hire for yet

Every department maps to real agents in the live catalog — not a roadmap.

Sales

Prospect research, buyer shortlists, warm-intro paths — workers that fill the top of your funnel while you sell.

Live now: Buyer Scout · Lead Research Agent · Discovery Scout

Explore sales

Finance

A CFO brief in your inbox every week, risk flags included — from nothing more than your ledger.

Live now: Weekly CFO

Explore finance

Growth

Validate ideas and chase the growth loops you never get to — research that runs itself.

Live now: SaaS Idea Validator

Explore growth
Meet the workers

Multi-step workers with a public track record

Every public worker ships with a live scorecard — runs, success rate, honest denominators.

First public workers are landing

The public catalog opens with workers whose scorecards are earned in production — no staged numbers. You don't have to wait: spawn your own worker from a plain-English goal today.

Spawn your own today
You stay in control

Governed autonomy, not a black box

These are screenshots of the real product — the same views you get on every run.

A task run on moulds.ai parked at an approval gate: the step card reads "Waiting for your approval" with an Approve & continue button — nothing proceeds without the owner.

Approval gates

Nothing outbound moves without you. Runs park at a gate and wait for your explicit approve.

A completed moulds.ai task run: numbered steps in a vertical list, each with a plain-English label, status pill, duration, and expandable input/output.

Plain-English run steps

Every run is a step list you can read — status, duration, inputs and outputs per step.

The AI-calls trace inside a moulds.ai task step: model, tokens in/out, retries, latency and timestamps per call, with the exact prompt and response expandable below.

Per-call AI trace

Model, tokens, latency — and the exact prompt and response — for every single AI call.

Proof

Real artifacts from real runs

Hand-curated excerpts of actual agent output — anonymized, never invented.

weekly-cfo

Weekly CFO — owner brief for a demo ledger (excerpt)

Weekly CFO Report — Northwind Trading — 2026-W07 Executive summary: Northwind Trading saw a significant improvement this week (2026-W07), with net movement increasing by $6,600 to $5,000, up from -$1,600 in the prior…

discovery-scout

Discovery Scout — weekly prospect package (anonymized excerpt)

Prospect package — 5 researched, 5 grounded, 0 duplicates 1. Harborlight seasonal open-air market — fit 8/10 - Who: Daniel Okafor, Founder/CEO — well-known recurring market event, a major East-Coast city - Why it fits:…

For builders: publish agents, earn on every run

Make your agent composable, list it in the catalog, and get paid through Stripe when people put it to work.

For builders
Inviting a small group at a time

Apply for early access

Tell us a little about you and we'll send an invite as we open more spots. Free to try, no card needed.

No spam — we'll only email about your invite.