Developers

Let AI agents create images and videos with Solart

One API key works for the REST API, the MCP server and the Agent Skill. Agents create images and videos with your account credits, and failed runs are refunded just like on the web.

Create an API key Read the guide

Works with

Claude Code
Claude Desktop
Cursor
VS Code
Codex CLI

Your agent

Make me a photo of a girl in a white ao dai by a lotus pond, 3:4

solart · generate{ model: "gpt-image-2", aspectRatio: "3:4" }

Image created1 image · 3:4 · saved to your history18

Here it is. Want a different setting or a few more variations?

MCP serverAdd one config entry and your agent can browse models, upload reference images and create images and videos.Best for Claude, Cursor, VS Code Agent SkillA single SKILL.md file that teaches your agent the API, no MCP install needed.Best for agents that run commands REST APICall it over HTTP from your app, scripts or automation tools.Best for apps and workflows

Quick start

1Create an API key under Account > API keys, copy the slr_... key and keep it safe.
2Pick how to connect: add the MCP server to your agent, install the Agent Skill, or call the REST API directly.
3Ask your agent to create an image. Credits are charged at the model's price, see the AI models page.

Try it with curl

curl https://solart.app/api/v1/me \
  -H "Authorization: Bearer slr_..."

MCP server

Solart runs a remote MCP server over Streamable HTTP. Add it to your agent once and it knows how to browse models, upload reference images and create images and videos.

MCP URLhttps://solart.app/api/mcpSend your API key in the Authorization: Bearer slr_... header

claude mcp add --transport http solart https://solart.app/api/mcp \
  --header "Authorization: Bearer slr_..."

Any client that supports Streamable HTTP only needs the MCP URL and the Authorization header above. For stdio-only clients use the mcp-remote bridge as in the Claude Desktop config.

Available tools

get_accountEmail and credit balance
list_modelsEnabled models, credit prices, supported ratios and resolutions
list_templatesReady-made prompt templates
get_templateVariables a template needs
upload_imageStore a reference image from a URL or base64
generateCreate an image or video and wait up to 45 seconds for it
get_generationStatus and result links, can wait longer
list_generationsRecent creation history

Agent Skill

The skill is a SKILL.md file that teaches agents how to call the Solart API: pick a model, upload images, create and wait for results, handle errors. Use it when your agent can run commands (curl) but you would rather not install MCP.

mkdir -p ~/.claude/skills/solart
curl -fsSL https://solart.app/skill.md -o ~/.claude/skills/solart/SKILL.md

The skill reads the API key from the SOLART_API_KEY environment variable.

export SOLART_API_KEY="slr_..."

Claude.ai and Claude Desktop: download the zip, then go to Settings > Capabilities > Skills > Upload skill.

Download skill (.zip)View SKILL.md

REST API

Use the base URL below and send the Authorization header with every request. Requests and responses are JSON, result links are absolute URLs.

Base URLhttps://solart.app/api/v1

Endpoints

GET/meAccount and credit balance
GET/modelsModels, prices and capabilities (filter with ?type=image|video)
GET/templatesTemplate list
GET/templates/{slug}Template details and variables
POST/uploadsUpload a reference image: multipart file, JSON url or base64 data
GET/generationsCreation history, paginate with before
POST/generationsCreate an image or video. ?wait=55 to wait for the result
GET/generations/{id}Status and result. ?wait=55 to wait
DELETE/generations/{id}Move to trash

Examples

curl -X POST "https://solart.app/api/v1/generations?wait=55" \
  -H "Authorization: Bearer slr_..." \
  -H "Content-Type: application/json" \
  -d '{
    "mode": "custom",
    "type": "image",
    "prompt": "A girl in a white ao dai by a lotus pond at sunrise",
    "params": { "aspectRatio": "3:4" }
  }'

Create an image from a prompt and wait for it in the same request.

Response

JSON

{
  "id": "8f0c...",
  "type": "image",
  "status": "succeeded",
  "outputs": [
    {
      "type": "image",
      "url": "https://solart.app/api/files/...",
      "downloadUrl": "https://solart.app/api/files/...?download=1"
    }
  ],
  "error": null,
  "creditCost": 24,
  "refunded": false,
  "filesExpireAt": "2026-10-13T09:00:00.000Z",
  "webUrl": "https://solart.app/generations/8f0c..."
}

Without wait the run is returned right away with status queued. Call GET /generations/:id?wait=55 until status is succeeded or failed. Images usually finish in 10 to 60 seconds, videos take a few minutes. Result files are deleted after filesExpireAt, download them if you need to keep them.

Limits and errors

Up to 20 creations per minute per account, with the same pending-run limit as on the web.

401Missing, invalid or revoked API key (code invalid_api_key).
402Not enough credits, top up on the Pricing page.
422Invalid parameters: ratio, resolution or duration not supported by the model, or missing template variables. Read error and details.
429Rate limit or pending-run limit reached, wait for the Retry-After header.

status failed: the run failed and its credits were refunded (refunded: true).

Agents can discover everything through /llms.txt and openapi.json.

curl -X POST "https://solart.app/api/v1/generations?wait=55" \ -H "Authorization: Bearer slr_..." \ -H "Content-Type: application/json" \ -d '{ "mode": "custom", "type": "image", "prompt": "A girl in a white ao dai by a lotus pond at sunrise", "params": { "aspectRatio": "3:4" } }'

{ "id": "8f0c...", "type": "image", "status": "succeeded", "outputs": [ { "type": "image", "url": "https://solart.app/api/files/...", "downloadUrl": "https://solart.app/api/files/...?download=1" } ], "error": null, "creditCost": 24, "refunded": false, "filesExpireAt": "2026-10-13T09:00:00.000Z", "webUrl": "https://solart.app/generations/8f0c..." }