Reduce your AI costs without rewriting code.

Access to the best open-source models, with a monthly plan and predictable billing, without paying per token.

CREATE API KEYSEE MODELS
A dark ocean wave cresting over a rocky shore, its spray lit in warm backlight against a slate-blue sky.

Wherever your company uses AI, Stac runs:in your product, your code and your agents.

Product

AI in your product, without a per-token bill

Chat, search, summarization and extraction running on open models, with the same API as OpenAI. Each new user stops meaning more tokens on the bill.

client.ts
 import OpenAI from "openai";   const client = new OpenAI({-  apiKey: process.env.OPENAI_API_KEY,+  baseURL: "https://api.trystac.com/v1",+  apiKey: process.env.STAC_API_KEY, });   const res = await client.chat.completions.create({   model: "kimi-k3",   messages: [{ role: "user", content: ticket }], });

Agents

Built for agents

An agent makes dozens of calls to deliver a single task. On Stac, you don't pay for each one of them by the token.

live agent sessions
  • Kimi K3Moonshot AI
    Session complete
  • GLM 5.3 FlashZ.ai
    Session complete
  • Qwen 3.8 27B (High)Alibaba
    Searching the web
  • Qwen 3.5 9BAlibaba
    Session complete
  • Qwen 3.8 27B (Low)Alibaba
    Running bash
  • Kimi K3Moonshot AI
    Session complete
api.trystac.com/v1cost per token $0.00

Coding

Your coding agent, on Stac

Use Stac in Claude Code and in OpenAI-compatible tools. Code all day without watching the cost climb with every prompt.

Automations

Automations that scale without scaling the bill

Connect Stac to n8n, Make or any workflow with an API key. Classify, extract and respond at volume.

localhost:5678/workflow/support-triage
n8n workflow: a webhook triggers an AI Agent whose chat model is the Stac Chat Model node (qwen/qwen3.5-9b), with Redis memory and a Supabase tool, then an If branch that runs JavaScript and sends a Gmail message.

Easy to integrate. Predictable to pay.

Stac handles the models, the infrastructure, and the billing, so your company just uses AI.

Per-request billing

Your plan covers a volume of requests. Beyond that, you pay per request, never per token.

Your data is yours

Nothing you send is used to train models.

Swap the URL, it just works

The same OpenAI API your code already uses.

Fast and stable

Low latency and stability for real-world use, not just demos.

Open-Source Models

Qwen and other leading open models, ready to use or fine-tuned on your own data.

Always up to date

A better open model comes out, it lands on Stac. You switch by changing one parameter.

From sign-up to first response in minutes

No GPU to rent, no server to spin up, no model to download.

  1. 01

    Create your account

    Sign up and pick the plan with the model and capacity your product needs.

  2. 02

    Generate an API key

    Create a key for your stack in the dashboard. One key, one endpoint.

  3. 03

    Swap the base URL

    Compatible with the OpenAI and Anthropic SDKs: point the base URL at Stac and your code stays the same.

Frequently asked questions