Your agent.
A direct path.

Find a model, check the published price and leave with a plan your coding agent can build from. Planning needs no account or token.

Prepare an integration Read llms.txt

SDKnpm · prodia

npm install prodia

Official TypeScript client for the Prodia API. Set PRODIA_TOKEN on your server.

Docs for agentsllms.txt

https://docs.prodia.com/llms.txt

Every job type and reference page as Markdown, for any coding agent.

Paste into Claude Code in your project:

Read https://prodia.com/llms.txt and https://docs.prodia.com/llms.txt, then add Prodia video generation to this project with the prodia npm package. Read the token from PRODIA_TOKEN and show me the expected cost before running any job.

Paste into Codex, or add to AGENTS.md:

Read https://prodia.com/llms.txt and https://docs.prodia.com/llms.txt, then add Prodia video generation to this project with the prodia npm package. Read the token from PRODIA_TOKEN and show me the expected cost before running any job.

Add https://docs.prodia.com/llms.txt under Cursor Settings → Indexing & Docs, then ask:

Using the Prodia docs, add Prodia video generation to this project with the prodia npm package. Read the token from PRODIA_TOKEN and show me the expected cost before running any job.

Any agent that can fetch URLs can plan with the HTTP API, no key needed:

Read https://prodia.com/llms.txt, prepare a plan with POST https://prodia.com/api/agent-plan, then implement it with the prodia npm package using PRODIA_TOKEN.

Two places to start.

Curated presets with fixed settings, not a ranking of the full catalogue. For other models, inputs or resolutions, use the docs index

text-to-video

MiniMax H3 Fast

A starting point for video with synchronized audio.

$0.2793 per 6-second requested clip

Job type: inference.minimax.h3.fast.txt2vid.v1
Preset: {"duration":6,"aspect_ratio":"16:9","resolution":"768P"}
Returns: video/mp4
Limits: 768P canvas class; exact dimensions depend on aspect ratio. Duration (4–15 s) rounds up to the model frame grid.

Specification Model Explorer

text-to-image

FLUX.2 Klein 4B

A low-cost starting point for image generation.

$0.003 per 1024 × 1024 image

Job type: inference.flux-2.klein.4b.txt2img.v1
Preset: {"width":1024,"height":1024}
Returns: image/jpeg
Limits: Text-only input for this preset. Other sizes or image inputs change the request and may change the price.

Specification Model Explorer

Published rates observed 10 November 2026 on the Prodia pricing page. They are not live quotes: recheck before every execution. Estimates exclude subscription fees, taxes, retries and negotiated discounts.

Prepare a plan.

Set the workload, check it against a budget and copy the brief into your coding tool. Nothing runs and nothing is charged.

Integration brief

Plan JSON

No generation executed. The request template produces one output when you run it with a valid Prodia token.

Plan over HTTP.

POST /api/agent-plan with JSON. No authentication. It returns a reference estimate, a budget check, one request template, sources and the credential handoff. It never runs inference, creates an account or charges anything.

Contract: openapi.json. Errors return HTTP 400, 413 or 415 with an error message.

cURL

curl -sS https://prodia.com/api/agent-plan \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "h3-fast",
    "quantity": 1000,
    "budget_usd": 300,
    "prompt": "An aerial camera drifts over an alpine ridge at sunrise, clouds rolling through the valley."
  }'

From plan to first result.

Your agent prepares the integration. Your account supplies the credentials and the budget.

Check the model and priceRead the current specification and published pricing, and show the configuration and expected cost to the user.
Connect your Prodia accountCreate a token in the API dashboard (requires a Pro subscription) and set PRODIA_TOKEN in your server environment. Never paste a token into this site.
Authorize one testApproval for a plan is not approval for the whole workload. Run one billable job first.
Implement on the serverFollow the inference reference; use async jobs for long work, bound retries, honor Retry-After and check job status before retrying an ambiguous timeout.
Check the output, then scaleConfirm quality and capture actual cost with the price flow, which runs a billable job.

This site offers HTML, plain text, JSON and an OpenAPI planning endpoint. It does not offer a remote MCP server, automatic signup, token provisioning, payment setup or live inference.