Coasty
A cloud-hosted AI agent platform that lets you automate desktop and browser tasks by describing what you want done, no infrastructure to manage.
🔗 Visit CoastyDescription
If you've ever wished you could just tell a computer 'click here, fill this in, then check that page' instead of writing a script for it, that's the gap Coasty tries to close. It's a cloud service where AI agents look at a screenshot of a screen, figure out where to click or type, and carry out multi-step tasks on your behalf — book something, fill a form, monitor a page — without you needing a server or dev-ops setup of your own.
Coasty exposes this through an API-first design across six core surfaces, taking screenshot input and returning precise coordinate outputs so agents can act reliably on real UIs. It provisions Linux and Windows VMs on demand, supports stateful multi-step workflows that persist session memory across steps, and can run unattended via cron-style scheduling and webhooks. Pricing is usage-based (credits), and it supports bring-your-own-model API keys so you pay your own LLM provider directly rather than a markup — Coasty only charges for its own managed model calls at $0.05 each. There's a free sandbox to test flows before committing to a paid tier.
💬 Our review
The short version: if you need to automate real UI interactions — the kind of task that's brittle with a scraping script and overkill for a full RPA suite — Coasty's screenshot-in, coordinates-out approach and bring-your-own-key pricing make it a genuinely cost-conscious way to get an AI agent driving a browser or desktop for you.
Against a full RPA platform like UiPath, Coasty is far lighter to set up and doesn't require enterprise licensing, but it also doesn't have UiPath's decades of enterprise workflow tooling. Against building directly on a raw computer-use API from a model provider, Coasty saves you the VM provisioning, session-state, and scheduling plumbing — that's the actual value here, not the AI itself. The free sandbox and BYO-key model make the entry cost low, though at $19-99/month plus per-call model costs, a team running agents constantly should model the credit math before committing to the Unlimited tier.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Essai gratuit sans CB. Starter 19$/mois (200 crédits), Plus 50$/mois (600 crédits), Unlimited 99$/mois. Appels modèle gérés 0,05$/appel ; clé LLM perso acceptée (bring-your-own-key, sans marge).
Pros
Approche screenshot → coordonnées fiable sur UI réelles
Bring-your-own-key sans marge sur l'inférence
Sandbox gratuite + VM Linux/Windows à la demande
Cons
Moins mature qu'une plateforme RPA enterprise (UiPath)
Coût des crédits à surveiller à l'usage intensif
Pas de couche low-code visuelle pour non-devs