Stac
Stac provides a drop-in OpenAI-compatible API endpoint that replaces per-token AI billing with a flat per-request fee starting at US$20 per thousand responses. The platform routes requests to frontier open-source models like Kimi K3, GLM 5.3 Flash, and Qwen variants, and includes built-in agent session management for long-running workflows. Developers switch by changing only the baseURL and API key in their existing code.
- Artificial Intelligence
- Developer Tools
- AI Agents
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Companies face unpredictable and rapidly escalating AI costs due to per-token pricing models from major AI providers. Rewriting existing code to switch providers or optimize token usage is time-consuming and risky, and the cost burden multiplies across production environments, embedded features, and autonomous agents.
Solution
Stac offers a drop-in OpenAI-compatible API endpoint that eliminates per-token billing entirely. Developers change their client's baseURL and API key—a single line of code—and immediately gain access to a portfolio of frontier models including Kimi K3, GLM 5.3 Flash, and multiple Qwen variants. Stac handles routing, session management, and execution for both synchronous calls and long-running agent workflows. The service presents a zero per-token cost model, replacing usage-based pricing with a fixed-fee approach that makes AI spend predictable.
Target Audience
Stac targets engineering teams and product organizations looking to integrate AI into production applications and agent workloads without the cost uncertainty of per-token billing.
Features
- Drop-in compatibility with the OpenAI SDK via a single baseURL and API key change
- Zero per-token pricing model, displayed directly in the dashboard as $0.00 per token
- Access to multiple frontier models from Moonshot AI, Z.ai, and Alibaba, including Kimi K3 and Qwen 3.8 27B
- Agent session management with live session tracking and automatic tool execution (bash, web search)
- Support for both low and high compute configurations of the same model to match performance needs
- Unified API endpoint (api.trystac.com/v1) for all models, enabling model switching without code changes