Skip to content

Call any model with one API

One OpenAI-compatible API and one key for Claude, GPT, Gemini, Grok and more: list models with prices, call them from the CLI, any OpenAI SDK, or MCP.

Dexto serves models from Anthropic, OpenAI, Google, xAI and others behind one OpenAI-compatible endpoint, https://api.dexto.ai/v1/chat/completions. One developer key reaches all of them, and every call is billed to your Cloud project’s credits. Switching models means changing one string.

List the models and their prices, then run one:

Terminal window
dexto cloud models list --type text
dexto cloud models get anthropic/claude-sonnet-5-5
dexto cloud models run xai/grok-4.7 "Explain idempotency keys in two sentences."

run takes the prompt as arguments or on stdin, and --system sets a system prompt:

Terminal window
cat CHANGELOG.md | dexto cloud models run anthropic/claude-sonnet-5-5 \
--system "Summarize for a release note. Five bullets at most."

With --json, run prints the whole chat completion, including usage.

Point the SDK’s base URL at Dexto and use a developer key:

import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.dexto.ai/v1",
apiKey: process.env.DEXTO_API_KEY,
});
const reply = await client.chat.completions.create({
model: "anthropic/claude-sonnet-5-5",
messages: [{ role: "user", content: "Explain idempotency keys in two sentences." }],
});

Streaming, tool calls, reasoning effort, and image, PDF and audio inputs work where the model supports them.

Terminal window
curl https://api.dexto.ai/v1/chat/completions \
-H "Authorization: Bearer $DEXTO_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "xai/grok-4.7",
"messages": [{"role":"user","content":"Explain idempotency keys in two sentences."}]
}'

GET /v1/models lists every model id with its context window and the reasoning_efforts it takes. See the Models API.

dexto_cloud_models_list and dexto_cloud_models_get read the catalog, and dexto_cloud_models_run runs a model in your Cloud project. A run is at the act level, so your client asks you first.

Each call is charged per token at the model’s price, which dexto cloud models list shows. The model catalog lists every model.