We build ChatGPT apps and production AI agents

We design, ship, and scale conversation-first experiences: Apps in ChatGPT (Apps SDK, preview), AgentKit workflows, Responses API tools (web+file search, computer use), and Realtime voice with SIP and image input—wired through MCP connectors

Our story

We started this studio to remove the friction teams face when turning great prompts into shippable, scalable agents. Instead of stitching vendors, we build on OpenAI’s newest stack so your customers get the experience right where they already chat.

Inside ChatGPT, we use the Apps SDK to render interactive app surfaces; on your site, we run the same logic with AgentKit and the Responses API tools for web search, file search, and computer use—so answers are fresh, grounded, and actionable.

For voice, we ship gpt-realtime agents with SIP calling, image input, and remote MCP access to your systems. We keep data private by default—OpenAI states it does not train on your business/API data—and we can prepare you for Instant Checkout via the Agentic Commerce Protocol.

It starts at the top

Results over process

We optimize for measurable wins, instrumenting agents with datasets, grading, and trace-based evals so we can prove improvements—not just ship prompts.

Embrace Accountability

Every project has clear KPIs and a visible decision trail (tools called, branches taken) using AgentKit’s visual workflows and evals.

Authentic Communication

We’re transparent about capabilities, limitations, and data handling. By default, OpenAI does not use your business data for training; we design with privacy and compliance in mind.

We are a remote team

We are lucky to be led by fearless leaders who constantly push Bonsa forward.

Antreas

Co-Founder

Athos