Ask an AI agent to read 300 sales calls and it works like one very patient person: it opens them one at a time. Everything it reads stays in its memory, so every next step is slower and costs more than the last one. By call 200, it re-reads a small library before it takes its next note.
Today we’re launching Ploy AI Gateway, so Ploy can call AI models directly and complete bulk tasks up to 8x faster and 20x cheaper than Claude Code and Codex. Ploy can also build AI-powered mini apps on your Ploy Site, no API keys needed.
AI workflows that run in parallel

Ploy gives each item (a meeting recording, a blog post, an image) its own AI workflow, runs them in parallel, up to hundreds at a time, and gathers results.
If you’ve built AI workflows in n8n or Zapier, this will look familiar, minus the building. Ploy builds the same kind of workflow from one prompt, picks the model for each step, and runs it on your Ploy credits.
AI on your website, too
Your published site can call the same models through Ploy AI Gateway. Visitors get live AI features, and you pay for them with the same Ploy credits.

These apps are very cheap to operate: a chatbot that reads your site answers 1,000 questions for under $1, and a decision model routes 14,000 visitors to the right page for $1. You set the spend limit for your sites on the Usage screen, so a busy day never surprises you.
The receipts: 300+ sales calls in 4.5 minutes
We gave Ploy, Claude Code, and Codex the same job: read 300+ of our recorded sales calls and rank the use cases prospects love most, with call counts and word-for-word quotes.
Ploy finished in 4.5 minutes for $2.87. The other runs took 17 to 39 minutes and cost $19 to $65, even with subagents.

| Ploy | Claude Code | Codex | |
|---|---|---|---|
| Time | 4.5 min | 17 min | 39 min |
| Total cost | $2.87 | $65 (est.) | $19.40 (est.) |
| Read every call in full | Yes | Yes | Yes |
| Quotes found word for word | 18 of 18 | 23 of 25 | 48 of 48 |
That’s up to 8x faster and 20x cheaper. Every run agreed on the top 3 use cases: SEO and AI search, account-based pages, and website builds and migrations.
When AI agents do the work, they re-read their instructions, the whole conversation, and every call they have already read, on every single step. An AI workflow reads one call and a few lines of instructions. Ploy’s 362 AI workflows ran on a smaller, faster model, and the whole batch finished in about 33 seconds.
Real-world usage of Ploy AI Gateway
Time measures the delay between the initial prompt and the final verified outcome, and costs include Ploy’s own work, not only the model calls.
| Job | Scale | Time | Total cost |
|---|---|---|---|
| Translate a whole site | Every page into 35 languages | 6 min | $1.51 |
| Fix missing alt text | 1,271 images sorted, 360 described | 12 min | $3.60 |
| Update training videos | 20 videos, 177 lines re-voiced | 9 min | $2.21 |
| Build lesson audio | 573 clips across 13 lessons, every clip checked | 25 min | $5.71 |
| Screen research papers | 693 abstracts ranked, top 40 read closely | 2 min | $0.73 |
| Score website visitors | 240 visitors scored, 127 qualified, 85 researched, 12 outreach drafts | 13 min | $6.21 |
These aren’t just one-shot jobs:
- They check their own work. A training-video job transcribed each new voice line back and compared it to the script.
- They know what to skip. An alt-text job wrote 360 descriptions and left 892 images alone where nearby text already explains them.
- They run again tomorrow. Saved as a Ploybook, a visitor-scoring workflow scores new visitors every morning.
AI Gateway building blocks
Ploy has access to a range of AI models. From a single prompt, it can combine them into a workflow or add them to your site.
| Block | Good for | Rough cost |
|---|---|---|
| Language models | Summaries, drafts, rewrites, translation, pulling facts out of messy text | 0.01 to 0.4 cents per item |
| Decision models | Yes or no checks, tagging, routing, and scores, always one of your options | ~0.005 cents per call |
| Image generation | Ad variations, blog headers, product scenes, photo edits | 3 to 8 cents per image |
| Voice and music | Voiceovers, read-aloud pages, soundtracks for ads | 0.2 cents per voice line, 4 cents per 30-second track |
| Speech to text | Call notes and captions, with word timings and speaker labels | ~0.4 cents per minute |
| Embeddings and reranking | Search by meaning, finding similar items, grouping | ~2 cents per 1,000 pages to embed, 0.2 cents per reranked search |
| Web search | Current facts and pages Ploy has no link for | ~0.7 cents per search |
| Video | Short clips from a prompt or an image | ~30 cents for a 4-second clip |
See them in action here:
3-step AI workflow
Write a personalized Ploy pitch to any business.
- FetchCode, no AILoads the home page and keeps the text a visitor can read.GetsWebsite→MakesPage text
Fetch
Loads the home page and keeps the text a visitor can read.
- RankDecision modelScores how well each of Ploy's 12 growth recs fits this business, in one call.GetsPage textPloy's 12 recs→MakesTop 3 recs
Rank
Scores how well each of Ploy's 12 growth recs fits this business, in one call.
- PitchLanguage modelWrites a short, casual pitch with one bullet for each top rec.GetsPage textTop 3 recs→MakesPitch
Pitch
Writes a short, casual pitch with one bullet for each top rec.
Use AI Gateway in your coding agent
Download the skill Ploy follows and add it to any coding agent, just provide a Vercel AI Gateway key. The skill covers picking a model for each job, running items in parallel, checking the results, and keeping an eye on cost.
Download the AI Gateway skill
4 things to try today
Paste any of these into Ploy:
- “Summarize every sales call from last quarter and rank the use cases prospects love, with quotes”
- “Make an ad image for each of our top 20 products, in the style of our best ad”
- “Translate our blog into Spanish, French, and German”
- “Find every image on our site with no alt text, and write it only where it’s needed”
Ploy tests a small sample first, then runs the rest in parallel. Before a big job, it tells you the estimated cost and waits for your go-ahead.
Run your first AI workflow
Ask Ploy for the outcome. It builds the AI workflows, picks the models, and runs them in parallel.