Skip to main content
AirGraphic is built to be driven by an agent. Hand your assistant one of the prompts below and it will set itself up, call the API, and return a finished infographic or meme.

​
Install

Pick whichever path fits the agent you’re using.

​
Paste a prompt

The fastest path. Copy one of these into Claude, Claude Code, ChatGPT, Codex, OpenClaw, Hermes, or any other agent that can read URLs and call HTTP APIs. One-shot generation:
docs-example: agent-prompt-now
Permanent setup:
docs-example: agent-prompt-setup
The setup prompt points the agent at our llms.txt, which is the LLM-friendly index of every doc page in markdown form. Most modern agents will fetch it, derive the request shape from the OpenAPI spec, and be ready to call AirGraphic on the next user message.

​
Drop in the skill file

For agent hosts that load YAML-frontmatter skills (Claude projects, Cursor rules, custom agent SDKs), grab the canonical skill file: docs.airgraphic.io/agent-skill.md It contains the install, authentication, request shape, and trigger patterns in one document the agent can load.

​
Authentication

Create a key at airgraphic.io/app/api-keys and expose it to your agent. Most users set an environment variable:
Keys begin with ag_, inherit your workspace plan and rate limits, and can be revoked from the same page at any time.

​
Compatible agents

AirGraphic works with any agent that can read a URL and make an HTTP call. The skill has been tested with:
  • Claude and Claude Code
  • ChatGPT and OpenAI Codex
  • OpenClaw, Hermes, and other autonomous agent platforms
  • Cursor, Windsurf, and other agentic IDEs
  • Custom agents built on the Anthropic, OpenAI, or AI SDK clients
If your agent can fetch https://docs.airgraphic.io/llms.txt and send a Bearer header, it can use AirGraphic.

​
Request pattern

For infographics, the agent should POST /api/generate, then poll GET /api/jobs/{id} until the status is terminal.
Poll /api/jobs/{jobId} with the same bearer key every 2.5 seconds until the status is completed, failed, or refunded. For memes, call POST /api/meme with sourceUrl, sourceText, a JSON sourceImage data URI, or a multipart sourcePdf/sourcePptx/sourceImage upload. Optional hints include audienceNote, avoidNote, styleNotes, brandUrl, brandProfileId, and aspectRatio. Meme jobs cost 2 credits and use the same polling endpoint.

​
What an agent can ask for

Pick one primary source per request:
  • A public URL (article, YouTube, TikTok video/photo, Instagram post/Reel, LinkedIn, X, Reddit)
  • Pasted text
  • A PDF, PPTX, or image upload
Optional brand and output hints:
  • brandUrl — any public site; AirGraphic extracts fonts and palette
  • brandProfileId — a saved brand from the user’s workspace
  • aspectRatio — 9:16, 1:1, 4:5, 16:9
  • outputFormat — png, jpeg, webp
  • outcome — target surface (LinkedIn carousel, Instagram post, blog header, etc.)
  • styleNotes — short free-form styling guidance
  • referenceImage — for style matching or as an explicit visual asset
See the full schema in the Reference and in the OpenAPI spec.

​
Example agent conversations

Once the skill is loaded, these all work:
  • “Make me a LinkedIn carousel from this Substack post: [URL]”
  • “Turn this PDF into a 1:1 Instagram post in Acme’s brand”
  • “Generate a story-format infographic on Q3 trends using these stats: …”
  • “Make a non-promotional meme for Acme’s target audience from this URL: [URL]”
  • “Re-run that same generation but in Stripe’s brand colors”
The agent picks the right source field, fills in sensible defaults (outcome from the user’s intent, aspectRatio from the target surface), calls the API, polls until the image is ready, and returns the imageUrl.