What is the Lamina API?

Lamina is a creative API that orchestrates AI image, video, and try-on models into reusable apps. You call a single endpoint with a brief, and Lamina handles model selection, brand context, and output formatting — so you ship creative without stitching raw model APIs together.

Which AI models does Lamina support?

Lamina supports 15+ image, video, and try-on models across leading providers, including FLUX, Stable Diffusion, Imagen, Veo, Runway, Kling, and Lamina-trained brand models. New models are added continuously and exposed through the same app interface, so you don't rewrite code when a better model ships.

Does Lamina support MCP?

Yes. Connect the hosted MCP with npx add-mcp https://app.uselamina.ai/mcp/agent/v2 — it approves over OAuth, so there is no API key to manage. The v2 endpoint exposes seven task-level tools (plan, execute, status, cancel, upload_asset, brand_context, credits); the parent /mcp/agent is an optional 30-tool advanced surface, and both share one OAuth resource. Any MCP client works: Claude, Cursor, Windsurf, VS Code, Zed. Stdio-only clients bridge with npx -y mcp-remote https://app.uselamina.ai/mcp/agent/v2. The old self-hosted @uselamina/mcp package is retired.

How does the Lamina API work?

You define a creative app once — output type, models, brand context, and guardrails — then call it with a brief. Each call creates a run, which Lamina tracks, evaluates, and distributes. The same app can be invoked from code, MCP, or no-code surfaces, and outputs stay consistent across channels.

What can I generate with Lamina?

Lamina supports six output types: product images, lifestyle scenes, on-model try-on, short-form video, banners, and 3D mockups. Each output type is brand-aware, so colors, fonts, and product fidelity stay locked across runs.

Why use Lamina instead of calling models directly?

Lamina removes the glue work of raw model APIs: prompt engineering, brand consistency, model fallbacks, evaluation, asset distribution, and version pinning. You call apps — not models — so creative quality is reproducible and your code survives model upgrades.

How do I get started with Lamina?

Sign up for an API key, install the SDK or MCP server, and call your first app in under five minutes. The free tier includes generation credits to evaluate the platform end-to-end before upgrading.

Which languages and SDKs does Lamina support?

Lamina ships official SDKs for TypeScript and Python, a CLI (@uselamina/cli), a REST API, and a hosted MCP server for AI agents. The REST API is OpenAPI-spec'd, so you can generate a client for any language.

How is Lamina priced?

Lamina is priced per generation, with volume discounts and an enterprise tier that includes dedicated capacity, SSO, and custom brand model training. See the pricing page for current per-output rates.

Is Lamina production-ready?

Yes. Lamina is in production with brands like Myntra, SHEIN, and ITC, generating creative at scale with sub-12-second median latency. The platform offers SLA-backed availability, evaluation hooks, and audit logs for every run.

Which Lamina MCP endpoint should an agent use?

Use https://app.uselamina.ai/mcp/agent/v2 — the recommended task-level surface with seven tools: lamina_plan, lamina_execute, lamina_status, lamina_cancel, lamina_upload_asset, lamina_brand_context, and lamina_credits. The parent /mcp/agent is an optional 30-tool advanced surface for direct app, model, compose, and checkout control. Both endpoints share one OAuth resource, so a single approval covers either.

Does planning spend generation credits?

No. lamina_plan only returns a frozen, priced plan — it never dispatches generation. Credits are spent only after you approve and lamina_execute runs, capped by the maxCredits budget you set. Use lamina_credits to check the balance before planning.

On-brand video generation for agents

Add lamina to ChatGPT or Claude and start creating on-brand cover art, shorts, reels and videos. No wrangling with model quirks or prompt optimization for video or custom workflows

Get started in 3 easy steps

Connect the MCP→API key (SDK / REST)Docs

quickstart — 3 steps

1npx add-mcp https://app.uselamina.ai/mcp/agent/v2

2plan → approve the cost → execute

3→ 4 brand-locked assets in ~12s

example / 01 · product shootrequest + response

// Claude / Cursor — one MCP install
$

→ returns 4 brand-locked assets in ~12sdocs.uselamina.ai/apps/product_shoot

set up with your agent

Install, activate, and ship from your agent.

Paste the prompt into Claude Code, Codex, Cursor, or ChatGPT. It connects the hosted MCP over OAuth (no API key), points the agent at the Lamina guide, and ships your first asset — no models to wire up.

hand this to your agent

Set up Lamina and make me an on-brand product photo.

1. Connect the MCP — run `npx add-mcp https://app.uselamina.ai/mcp/agent/v2` and approve the OAuth prompt in your browser. No API key needed.
2. Learn the tools — read https://uselamina.ai/agents.md. (In Claude Code you can also run `npx @uselamina/skill` to install it as a skill.)
3. Plan first — call `lamina_credits` and `lamina_brand_context`, then `lamina_plan`. Show me the frozen steps and estimated cost and wait for my approval.
4. On approval, call `lamina_execute` (with a maxCredits ceiling), then poll `lamina_status` until it's done and show me the result. Planning alone never generates anything.

just connecting?npx add-mcp https://app.uselamina.ai/mcp/agent/v2MCP · OAuth, no key · code · @uselamina/sdk (api key)

works in · claude code · codex · chatgpt · cursor · windsurf · vs code · zedmachine-readable · /agents.md · llms.txt

the platform / 03

One pipeline. Four endpoints.

Every Lamina app — from product shoots to brand films — is built on the same four primitives. Compose them, or call the apps that already chain them for you.

base · app.uselamina.ai/v1
auth · bearer

→

01Create

POST/v1/content/create

↓

Pass a brief (or pick an app). Lamina routes it across the right models.

→

02Track

GET/v1/runs/{runId}

↓

Poll or stream a run — queued → composing → rendering → done — with progress.

→

03Evaluate

POST/v1/content/score

↓

Score each asset against the brand kit. Reject off-brand before they ship.

04Distribute

POST/v1/publishing/publish

↓

Publish to your connected channels — Shopify, socials, S3, or a webhook.

outputs / 04

What you can generate.

Six built-in apps cover most of the catalogue. Each one is a tuned pipeline — not a single model — and each one knows your brand.

see all 30+ apps →

apps.productShoot()

Product shoots

Studio-grade stills with controlled backgrounds, angles, and lighting.

lamina.runs.run("product_shoot", { brief, brand })

1:14:516:95 params

apps.verticalReel()

Vertical reels

Motion-first short video for Instagram, TikTok, YouTube Shorts.

lamina.runs.run("vertical_reel", { brief, brand })

9:165 params

apps.adVariant()

Ad variants

Localized, resized, on-brand display ads. Dozens from one brief.

lamina.runs.run("ad_variant", { brief, brand, sizes })

1:14:516:99:165 params

apps.virtualTryOn()

Virtual try-on

Garments on AI models across body types, poses, and settings.

lamina.runs.run("virtual_try_on", { garment_url, brand })

4:51:14 params

apps.campaignBanner()

Campaign banners

Hero banners and display ads matched to your brand system.

lamina.runs.run("campaign_banner", { brief, brand })

16:91:14:55 params

apps.brandFilm()

Brand films

Storyboard, generate, and edit short brand films end-to-end.

lamina.runs.run("brand_film", { brief, brand })

16:99:165 params

positioning / 05

Why not just call models?

Raw model endpoints are the easy part. What kills you is the orchestration — and the brand drift. Lamina is the layer between you and the chaos.

capability	raw model endpoints	with lamina
Brand consistency	×Manual prompt engineering per run	✓Brand kit attached once, enforced everywhere
Model routing	×You pick and manage each model	✓Auto-routed across 15+ models
Multi-ratio output	×Separate call per aspect ratio	✓One call, all ratios returned
Asset evaluation	×Manual review of every output	✓Automated brand-fit scoring with reasons
Distribution	×Build your own pipeline	✓S3, Shopify, Sanity, Drive built-in
SDK surface	×5 SDKs, 5 auth flows, 5 webhooks	✓One SDK. One webhook. One auth header.
Billing	×5 invoices, 5 quotas to reconcile	✓One bill. One quota. One dashboard.
Agent support	×Build tool wrappers yourself	✓14 MCP tools, connect in one command

Every output stays on your brand — across every model, ratio, and channel.

Brand consistency

×Manual prompt engineering per run

✓Brand kit attached once, enforced everywhere

Model routing

×You pick and manage each model

✓Auto-routed across 15+ models

Multi-ratio output

×Separate call per aspect ratio

✓One call, all ratios returned

Asset evaluation

×Manual review of every output

✓Automated brand-fit scoring with reasons

Distribution

×Build your own pipeline

✓S3, Shopify, Sanity, Drive built-in

SDK surface

×5 SDKs, 5 auth flows, 5 webhooks

✓One SDK. One webhook. One auth header.

Billing

×5 invoices, 5 quotas to reconcile

✓One bill. One quota. One dashboard.

Agent support

×Build tool wrappers yourself

✓14 MCP tools, connect in one command

Every output stays on your brand — across every model, ratio, and channel.

agents / 06

Built for AI agents.

Seven task-level tools at /mcp/agent/v2 — the recommended surface. An agent checks credits and brand context, plans, shows a frozen plan and cost, and only executes after explicit approval within a credit budget; planning never spends generation credits. A separate 30-tool advanced surface at /mcp/agent stays available (optional) for direct app, model, compose, and checkout control.

Structured Output

Typed JSON responses with Zod schemas. Perfect for agent parsing.

Idempotent & Streamable

Safe to retry. SSE for every long-running job. Agents stay informed.

Deterministic

Same brief + brand + seed produces the same output. Reproducible pipelines.

agent workflow · mcp v2

plan → approve → execute

// Recommended v2 flow: plan → approval → execute → deliver
// Works in Claude, Cursor, Windsurf, or any MCP client

// 1. Check credits + load brand context to ground the plan
await lamina_credits();
const brand = await lamina_brand_context();
// → { palette: [...], voice: "warm minimal", dos: [...], donts: [...] }

// 2. Plan — returns a frozen plan + cost. Does NOT generate anything.
const plan = await lamina_plan({
  brief: "linen tee on teak, north light, 35mm, 4 variants",
  brand: brand.id
});
// → { fingerprint, steps: [...], estimatedCredits: 8 }

// 3. Show the frozen steps + cost and get explicit approval, then
//    execute within a credit ceiling with an idempotency key
const run = await lamina_execute({
  fingerprint: plan.fingerprint, maxCredits: 10,
  idempotencyKey: "ss26-hero-01"
});

// 4. Poll until terminal (~25s waits; retry on timedOut: true)
const done = await lamina_status(run.id);
// → done.outputs[].url  ·  cancel any time with lamina_cancel(run.id)

Need direct app, model, or compose control? The optional https://app.uselamina.ai/mcp/agent endpoint exposes a 30-tool advanced surface. See the advanced tool reference ↗It shares the same OAuth resource. Start with the seven v2 tools above; reach for advanced only when you need low-level control.

works in · claude · cursor · windsurf · vs code · zed · raycast · n8n · zapierMCP or HTTP · install in one command

FAQ

Common questions about Lamina

Lamina is a creative API for orchestrating AI image, video, and try-on models behind a single, brand-aware interface. Below: how the API works, what it supports, and how it compares.

Basics

What is the Lamina API?
Lamina is a creative API that orchestrates AI image, video, and try-on models into reusable apps. You call a single endpoint with a brief, and Lamina handles model selection, brand context, and output formatting — so you ship creative without stitching raw model APIs together.
How does the Lamina API work?
You define a creative app once — output type, models, brand context, and guardrails — then call it with a brief. Each call creates a run, which Lamina tracks, evaluates, and distributes. The same app can be invoked from code, MCP, or no-code surfaces, and outputs stay consistent across channels.

Models

Which AI models does Lamina support?
Lamina supports 15+ image, video, and try-on models across leading providers, including FLUX, Stable Diffusion, Imagen, Veo, Runway, Kling, and Lamina-trained brand models. New models are added continuously and exposed through the same app interface, so you don't rewrite code when a better model ships.
What can I generate with Lamina?
Lamina supports six output types: product images, lifestyle scenes, on-model try-on, short-form video, banners, and 3D mockups. Each output type is brand-aware, so colors, fonts, and product fidelity stay locked across runs.

Integration

Does Lamina support MCP?
Yes. Connect the hosted MCP with npx add-mcp https://app.uselamina.ai/mcp/agent/v2 — it approves over OAuth, so there is no API key to manage. The v2 endpoint exposes seven task-level tools (plan, execute, status, cancel, upload_asset, brand_context, credits); the parent /mcp/agent is an optional 30-tool advanced surface, and both share one OAuth resource. Any MCP client works: Claude, Cursor, Windsurf, VS Code, Zed. Stdio-only clients bridge with npx -y mcp-remote https://app.uselamina.ai/mcp/agent/v2. The old self-hosted @uselamina/mcp package is retired.
Which languages and SDKs does Lamina support?
Lamina ships official SDKs for TypeScript and Python, a CLI (@uselamina/cli), a REST API, and a hosted MCP server for AI agents. The REST API is OpenAPI-spec'd, so you can generate a client for any language.

Comparison

Why use Lamina instead of calling models directly?
Lamina removes the glue work of raw model APIs: prompt engineering, brand consistency, model fallbacks, evaluation, asset distribution, and version pinning. You call apps — not models — so creative quality is reproducible and your code survives model upgrades.

Getting Started

How do I get started with Lamina?
Sign up for an API key, install the SDK or MCP server, and call your first app in under five minutes. The free tier includes generation credits to evaluate the platform end-to-end before upgrading.

Pricing

How is Lamina priced?
Lamina is priced per generation, with volume discounts and an enterprise tier that includes dedicated capacity, SSO, and custom brand model training. See the pricing page for current per-output rates.

Production

Is Lamina production-ready?
Yes. Lamina is in production with brands like Myntra, SHEIN, and ITC, generating creative at scale with sub-12-second median latency. The platform offers SLA-backed availability, evaluation hooks, and audit logs for every run.

MCP

Which Lamina MCP endpoint should an agent use?
Use https://app.uselamina.ai/mcp/agent/v2 — the recommended task-level surface with seven tools: lamina_plan, lamina_execute, lamina_status, lamina_cancel, lamina_upload_asset, lamina_brand_context, and lamina_credits. The parent /mcp/agent is an optional 30-tool advanced surface for direct app, model, compose, and checkout control. Both endpoints share one OAuth resource, so a single approval covers either.
Does planning spend generation credits?
No. lamina_plan only returns a frozen, priced plan — it never dispatches generation. Credits are spent only after you approve and lamina_execute runs, capped by the maxCredits budget you set. Use lamina_credits to check the balance before planning.

trust & scale

Already shipping at production scale.

The same API powering 500+ brand teams now powers your agents. SOC 2 certified, EU residency available, 99.95% uptime SLA on the production tier.

500+

teams using the API

4.1M

assets shipped (30d)

99.95%

API uptime

11.8s

p50 latency

96.4%

brand-fit pass rate

<2min

time to first asset

compliance

SOC 2 Type IIGDPREU residency99.95% SLASSO / SCIM

powering teams at

HomesteadMarlowAster

get started

Stop calling raw models.
Call agents.

Free tier — 100 generations, no card. Connect the hosted MCP, the CLI, or the SDK. Teach your coding agent the whole flow with `npx @uselamina/skill`. Full reference at docs.uselamina.ai.

Get API key →$ Read the docs Try in Claude ↗

open docsAPI referencebrowse 30+ appsjoin communitygithub