Use with AI

Let AI coding tools (Cursor, Claude Code, VS Code + Copilot/Cline, Windsurf, etc.) read this API documentation while writing your integration code.

Option 1 — Plain-text docs (no setup)

The guide pages and the API definition are served as machine-readable files:

/openapi/artguru-open-api.json — the OpenAPI 3.1 spec: every endpoint, parameter, and response shape. This is the file to give an AI that needs to write API calls.
/llms.txt — index of the guide pages (concepts, authentication, pricing, errors) as raw Markdown
/llms-full.txt — all guide pages in one file

The llms*.txt files cover the guides only; endpoint schemas live in the OpenAPI spec. Both llms*.txt files link to it, so pointing an AI at any one URL is enough.

Option 2 — MCP server

For richer integration (endpoint listing, per-endpoint schema lookup, optional real API invocation), add an MCP server backed by this site's OpenAPI spec. Add the following to your tool's MCP configuration (e.g. .cursor/mcp.json, claude mcp add, or your IDE's MCP settings):

{
  "mcpServers": {
    "artguru-api-docs": {
      "command": "npx",
      "args": ["-y", "@ivotoby/openapi-mcp-server@1.16.1"],
      "env": {
        "API_BASE_URL": "https://api.artguru.ai",
        "OPENAPI_SPEC_PATH": "https://docs.artguru.ai/openapi/artguru-open-api.json",
        "TOOLS_MODE": "dynamic"
      }
    }
  }
}

This exposes three tools to your AI assistant. The assistant discovers their input schemas automatically; the shapes are shown here so you can also call them by hand from an MCP inspector:

Tool	Arguments	Returns
`list-api-endpoints`	none	Every operation with method, path, summary, and tags
`get-api-endpoint-schema`	`endpoint` — the path only, e.g. `/api/v1/upscale/generate`	Request parameters, body schema (including `mode`, `resolution`), and responses
`invoke-api-endpoint`	`endpoint`, `method` (`GET`/`POST`…), `params` — sent as the JSON body for POST	The live API response

Example — ask for the Upscale Pro schema, then submit a task:

{ "name": "get-api-endpoint-schema",
  "arguments": { "endpoint": "/api/v1/upscale/generate" } }

{ "name": "invoke-api-endpoint",
  "arguments": {
    "endpoint": "/api/v1/upscale/generate",
    "method": "POST",
    "params": { "image": "https://example.com/photo.jpg", "resolution": "4k", "mode": "portrait" }
  } }

The schema returned for enhance, unblur, and upscale includes the mode parameter (general / graphics / portrait / text), so the assistant can choose it for you.

Node.js >= 18 must be installed.

Read-only by default

With the configuration above, the AI can only read the API documentation — it cannot call the Artguru API on your behalf. To allow real API calls (invoke-api-endpoint), additionally provide your key:

"env": {
  "API_HEADERS": "x-api-key:<your_api_key>"
}

Only do this if you want the AI to execute real, credit-consuming requests. Ask the assistant to read Pricing before estimating costs: rates depend on the endpoint, resolution, and mode. A credit-balance response alone does not show how many requests it covers.

Supply-chain note

@ivotoby/openapi-mcp-server is a community-maintained package. The version is pinned above on purpose — avoid @latest so that an unreviewed release cannot run on your machine automatically.

Option 1 — Plain-text docs (no setup)​

Option 2 — MCP server​

Option 1 — Plain-text docs (no setup)

Option 2 — MCP server