One AI video API across the keys you already pay

The short answer

1. One studio graph, as an API. Everything the app does is a route under /v1, and the same routes are the tools an agent gets over MCP. Rendering spends on the workspace’s own provider keys. Nothing else spends.
2. Your keys, not a second aggregator account. You already pay Google, Kling, fal, or whoever issued the credential. lotgen does not resell those seconds. The seat is the studio. The labs bill the clip.
3. A quote before money moves. The same render route with dry_run: true spends nothing and needs only read. The response carries total_usd and cap.
4. A lot you can review. A lot is an ordered list of shots. Each render is a take. Compare, mount a winner, share a review link. Re-run shot three without touching the other four.

Twelve generation keys lotgen.ai pastes: Google, BytePlus ARK, fal, evolink, Kling, OpenAI, MiniMax, Black Forest Labs, Alibaba, ElevenLabs, xAI and Runway. Luma and Sora are models on fal or evolink, not keys. BytePlus ARK is the console that issues the Seedance key, and the name to put on the invoice. The invitation form also lists Anthropic, which buys text and nothing this page prices.

What “AI video generation API multiple providers” means here

On this page, an AI video generation API for multiple providers is one workspace graph: lots, shots, takes, quotes, caps, review, export. You paste the keys you already pay. You pick a provider on the lot, or a model on the shot. You get a dollar figure before anything hits a lab. You review the take in the same place a person would.

That is a different job from buying a second account that happens to list many models.

A unified API for AI video models, in the sense search uses that phrase, is usually a marketplace catalog: one vendor key, their price list, their queue. fal and Replicate are good at that job. lotgen is the lot on top of the keys you already hold. The quote is the lab’s list rate on your credential, not a reseller row with a markup you cannot see.

Pricing says it in one sentence: you pay the labs for seconds, at their list rate, on your own keys, and you pay lotgen for the studio. There is nothing in between: no packs, no units, no percentage on your keys.

The Academy lesson bringing your own keys: a key belongs to a platform, not to a model. One Google key pays for Omni, Veo, Nano Banana, and Gemini drafts. A render with no working key for the provider fails. It does not quietly run on a lotgen-owned credential.

A lotgen workspace key starts lgk_. Four scopes, from /developers: read / write / render / admin. read is implied by the other three. write spends nothing. render is the scope that spends on your provider keys. The mint dialog, curl, SDK, rate-limit headers, webhooks, and OpenAPI stay on that page. This page does not copy them.

Pick a provider per lot, or per shot

default_provider on a lot is any provider the workspace holds a key for. The public example on /developers uses "fal". Creating the lot spends nothing.

A shot on that lot can pick a different model, and therefore a different key, than the lot default. The Academy lesson picking a model: listing providers, listing models, and quoting a shot are all free. Quote the shot you actually mean (references, duration, the model named) because a marketplace headline is a union no single model keeps. Render one shot on the candidate before you render the rest of the lot.

A platform can hold more than one key. Each has a name and an owner, one is the default, and a lot can be bound to a specific key. That is how client work lands on the client’s credential and the client’s bill.

Do not treat Luma or Sora as one more video key. They are models on fal or evolink. Sora is not the OpenAI key this product pastes (that key is stills and copy). OpenAI’s own Sora 2 list, for the model-on-marketplace case, is on how much AI video generation costs; this page does not redo it.

If a route is not on /developers or in the OpenAPI it links, it does not exist for this page. Same-prompt footage across keys is a compare still to come. It needs a real run first. This page does not invent it, and it does not claim “we ran 50 jobs across twelve keys.”

Quote and cap before spend

The product match for an agent loop, and for a technical buyer who already pays per-second rates, is a dollar figure in the response plus a cap the loop cannot blow through. A Veo channel without a pause can wipe a budget in hours. An agent makes that worse. Do not skip the quote.

The quote, from /developers: the same render route with dry_run: true. It spends nothing and needs only read. The public example response shape:

{ "dry_run": true, "total_usd": 0.42, "estimate": { }, "cap": { } }

total_usd is the quote. cap is the bound. Every route that renders takes dry_run. A dry run counts as a read, not a render, for rate limits (600 requests and 30 renders a minute per key). The $0.42 figure is the public docs example and the landing’s illustration of a 5-second Kling 3.0 take at 720p, quoted direct on a Kling key. It is a worked example, not a measured run.

Caps, from pricing and the Academy lesson spend caps: per workspace and per lot. The workspace cap is the month’s total. The lot cap is one project’s own budget. Either can refuse, and the refusal names which one did, what is already committed, what this render would add, and the headroom left. A queued render counts against the cap immediately. A cap is optional; clearing it is a single action. The argument for one here is that it is scoped to this tool, including to an agent doing exactly what it was told to do.

Unpriced models refuse under a cap rather than walking through as $0.00. Pricing still shows Wan 2.7 1080p as unpriced when checked. Believe the button.

What a second costs across the twelve keys is how much AI video generation costs (prices checked 2026-09-18 on that page). Paste your own spec into the AI video cost calculator (toggle “I’ll run this via API or MCP” if that is the job). Eight-second Kling vs Veo dollars: Kling vs Veo cost. Veo tiers: Veo 3 cost per second. Kling credits vs API Units: Kling pricing explained. This page does not copy those grids.

Same graph in the app, the SDK, and MCP

Anything you can do in the app, an agent can do over MCP: the same workspace, the same quotes, the same caps. MCP is one client of the same /v1 graph. It is not a second product.

Three doors, one graph:

• The app. Six studios plan a lot. From there it is ordinary shots, takes, and winners. Review links, export pack, spend caps, usage.
• The SDK. npm i lotgen. The client the MCP server itself is built on, so it cannot drift from the tools. Typed end to end. The public snippet lives on /developers. This page does not paste the five-call tutorial.
• MCP. Claude, Cursor, Claude Code, or a script. Public URL: https://mcp.lotgen.ai/mcp. Without a key it answers 401, which is the endpoint up, not an outage. Sign-in vs connection key is on Connecting Claude. This page does not rewrite those connect steps.

A key that lacks the scope a route needs answers 403 insufficient_scope naming it. A brand-bound key (the Partner add-on) sees one client’s work. Usage rolls up by brand and by key.

There is no tool catalog on this page. Connecting Claude is explicit: the tool list is read from the running server so the page cannot list a tool that was removed or miss one that was added. The landing prints the live count the same way.

Not fal, not Replicate, not a second Kling MCP

Those are model catalogs. This is the lot.

fal, Replicate, and a Kling-only SDK (or a hobby kling mcp wrapper) are the right tool when you want one vendor’s models and you will hold that vendor’s auth yourself. Search for “ai video generation api multiple providers” currently ranks aggregator sites; that is the gap this page is for, and it is not filled by listing more models.

lotgen is quote, cap, review, and export on the workspace’s own provider keys. Switching labs is picking a key on the lot, not installing a second vendor MCP. Kling on lotgen is the Kling key you already pay, or Kling as a model on fal/evolink. Veo is the Google key. The quote is that provider’s list rate, on your key, in total_usd.

What it costs

This is not a separate SKU. Pricing: MCP on every plan. Claude over MCP is the same price as the UI.

A dry-run quote is free. Planning a lot is free. The seat is the product. The labs are the other line on the card. The live cost pages carry the labs’ side: how much AI video generation costs and the AI video cost calculator.

Request access

lotgen.ai is invite-only beta. The door is the invitation form. Pick Developer or agent builder (“You want the API or MCP more than the studio.”). The form does not create an account and it charges nothing. A person reads it. The providers you already pay for is the part that shapes the answer. If it is a fit, the invitation arrives by email and creates the workspace when you accept it.

Request an invitationSeats and pricingYou want the API or MCP more than the studio.

Or start today: Solo is $99 a month and checkout creates the workspace. Same graph, same keys, no queue. Seats and pricing. The API/MCP contract, once you have a key: /developers.

Product facts checked on 2026-09-18: how to re-check yourself

1. The contract: /developers. default_provider, dry_run to total_usd and cap, the lgk_ prefix, the four scopes, the per-key limits, the Partner add-on price.
2. The seats and MCP on every plan: Pricing. If MCP ever becomes an add-on, this page needs rewriting.
3. The keys: the invitation form’s checkbox list at /request and the named list on how much AI video generation costs. Luma and Sora are models on those lists, not keys.
4. The endpoint: https://mcp.lotgen.ai/mcp without a key answers 401. A 404 or a 5xx is a different story.
5. The lessons this page leans on: bringing your own keys, picking a model, spend caps, what a lot is, connecting Claude.

Related on this site: the MCP how-to, generate AI video from Claude Code.

Request an invitationSeats and pricingA dollar quote on your own keys before an agent runs a lot.

FAQ