APIMart

APIMart

APIMart is a single gateway for GPT, Claude, Gemini, Grok, and dozens of image, video, audio, and music models. Pick the upstream model in the request instead of switching services per vendor.

Capabilities

• Account — check API key balance or whole-account credits before long media runs.
• Chat — OpenAI-style chat completions across text models.
• Images — async image generation; jobs return a task id until URLs are ready.
• Videos — async video generation; same task pattern as images.
• Uploads — host reference images when a flow needs a URL instead of inline bytes.
• Tasks — use the wait-for-task tool after async submits, or a one-shot status read when you only need a snapshot.
• Extended surface — the generic API request tool covers Claude Messages, Whisper, TTS, Suno/Flow music, Midjourney, moderation, and other documented routes when no dedicated tool exists.

Workflow

1. Validate — check balance before expensive image or video batches.
2. Text — chat completions with streaming off for full JSON in agents.
3. Reference media — upload images first when a flow requires a hosted URL.
4. Media — submit with wait-for-completion, or capture the task id and call the wait-for-task tool until URLs appear in the result (avoid manual poll loops).
5. Long tail — use the API request tool for routes without a dedicated tool (music tasks, Midjourney imagine, and similar).

Gotchas

• Async media — image and video jobs return a task id; use wait-for-task or wait-for-completion for final URLs. Links often expire in 24–72 hours.
• Billing — per-task cost and credits appear on completed tasks; usage history and invoices are dashboard-only.
• Model IDs — use APIMart model slugs (for example gpt-4o, sora-2, gpt-image-1-official); pick from the APIMart model catalog, not vendor-native ids unless documented as aliases.
• Insufficient balance — payment required; top up at APIMart before retrying.
• Chat streaming — MCP tools return complete JSON; disable streaming on chat requests.