Vercel AI Gateway for AI agents
Agents read Vercel AI Gateway at once. Changes and sends are held for a person's approval by default.
About Vercel AI Gateway
Access Vercel AI Gateway's model discovery, inference, generation, and usage APIs through a unified AI model gateway.
What an agent can do in Vercel AI Gateway
Every Vercel AI Gateway action is classed by the platform as a read, a change or a send. Reads run at once. Changes and sends are held for a person by default.
Runs at once (reads)
Looks things up in Vercel AI Gateway without asking first. 8 actions.
- Get Credit BalanceReturn the connected team's remaining AI Gateway credit balance and lifetime credit usage in USD.
- Get GenerationLook up cost, token usage, provider routing, latency, and finish details for one completed Vercel AI Gateway generation.
- Get ModelReturn detailed capabilities, modalities, limits, tags, and pricing for one exact Gateway model ID.
- Get Model EndpointsCompare provider endpoints serving one Vercel AI Gateway model, including limits, pricing, caching, availability, throughput, and latency.
- Get Spend ReportQuery aggregated Vercel AI Gateway spend and token usage. This endpoint requires a Pro or Enterprise plan, and each query costs $5 per 1,000 queries.
- List Model CatalogList AI Gateway model-to-provider routes and capability tags, with bounded local filtering by model text, provider, or tag.
- List TypeSafe ModelsList evaluation models available through AI Gateway's TypeSafe-compatible API, including each accepted model name, description, and release date.
- Search ModelsSearch the AI Gateway model catalog by text, model type, and tags, returning capability and pricing metadata for matching models.
Held for approval by default (changes and sends)
Held for a person's approval by default. The approval is recorded and the action is sent once, never twice, with 10 seconds to undo. 8 actions.
- Edit ImagesCreate edited images from up to 16 HTTPS or base64 data-URL source images and a text instruction. Uses the non-streaming JSON image-edit route only.
- Evaluate StateEvaluate one shared state against one or more typed choice, score, or boolean questions using an AI Gateway evaluation model. Returns structured answers, token usage, and gateway metadata rather than
- Generate ImagesGenerate one or more images from a text prompt through an image-only Vercel AI Gateway model. Returns a completed non-streaming JSON response.
- Generate SpeechGenerate spoken audio from text through a Vercel AI Gateway speech model. Returns completed base64 audio in JSON; streaming output is not supported.
- Rerank Object DocumentsRank structured object documents by relevance through the Cohere v1 route, using selected rank fields and optionally returning each source document.
- Rerank Text DocumentsRerank string documents through the current Cohere v2 compatibility endpoint. Unlike the v1 action, this action accepts text strings only and does not support object documents, rank_fields, or return_
- Transcribe AudioTranscribe base64-encoded recorded audio through a Vercel AI Gateway transcription model. Returns one completed JSON transcript, not a live stream.
- Evaluate with TypeSafe APIEvaluate one shared state through AI Gateway's TypeSafe-compatible System One API. Use this compatibility action for TypeSafe noul, choice, and score request shapes; use Evaluate State for new integra
Changes wait in Needs you
Changes and sends in Vercel AI Gateway are held for a person's approval by default. The approval is recorded, and the action is sent once, never twice, with 10 seconds to undo.
Related connectors
Vercel AI Gateway questions
Can an agent change data in Vercel AI Gateway on its own?
Not by default. Changes and sends in Vercel AI Gateway are held for a person's approval. The approval is recorded, and the action is sent once, never twice, with 10 seconds to undo.
What can an agent read in Vercel AI Gateway?
This connector lists 8 read actions. Reads run at once, without waiting for approval.
How does Vercel AI Gateway connect?
It connects with API key. Connectors are built and maintained by third-party developers; availability is confirmed during scoping.