What is OpenAI Agent Kit?

OpenAI Agent Kit is a developer toolkit for building agents that call the OpenAI API directly. It includes SDKs for common languages (Python, TypeScript), tool-calling primitives, memory patterns, and examples. It’s the code-level answer to ‘how do I build an agent?’, in contrast to Workspace Agents, which is the no-code, team-scoped product inside ChatGPT.

Is OpenAI Agent Kit the same as Workspace Agents?

No. They’re different products for different buyers. Agent Kit is for developers building custom agents against the OpenAI API, and you control the infrastructure, the runtime, and the deployment. Workspace Agents is a managed product inside ChatGPT Business and Enterprise for teams that want agents without building their own runtime. Same company, different layer.

When should I use Agent Kit instead of Workspace Agents?

Use Agent Kit when you’re shipping an agent inside your own product (not for internal team use), when you need custom infrastructure or on-prem deployment, or when the agent needs integrations Workspace Agents doesn’t support. Use Workspace Agents for internal team workflows where the goal is to save your team’s time, rather than build a product feature.

Yes. It’s common for a company to use Workspace Agents internally (sales, support, ops automation) and Agent Kit externally (user-facing AI features in their product). Different audiences, different requirements.

Does Agent Kit cost more than Workspace Agents?

Different billing models. Agent Kit uses standard OpenAI API pricing (per-token, per-model). Workspace Agents uses credit-based workspace pricing on top of your Business/Enterprise plan. For high-volume agents, Agent Kit can be cheaper because you pay only for raw model usage without the workspace credit markup. For low-volume internal agents, Workspace Agents is usually cheaper because you don’t have to pay for infrastructure.

Do I need both for my team?

Almost never. If your team doesn’t ship software to customers, you just need Workspace Agents. If you do, you might want both, but start with Workspace Agents and add Agent Kit only if a specific product feature requires it.

Guide · Updated April 23, 2026

OpenAI Agent Kit vs Workspace Agents

Two OpenAI products with overlapping names and very different audiences. This page clears up the confusion and helps you pick the right one for your situation.

Side-by-side

Dimension	OpenAI Agent Kit	Workspace Agents
Audience	Developers shipping product features	Teams automating internal workflows
Interface	SDK, code-level	ChatGPT UI, no-code
Runtime	Your infrastructure	OpenAI-hosted in ChatGPT workspace
Deployment	Anywhere you deploy code	Inside your ChatGPT tenant
Pricing	OpenAI API per-token	Workspace credits (on top of Business/Enterprise seat)
Customization	Full control: model, memory, tools, routing	Opinionated, prompt-driven authoring
Admin governance	You build it	Built-in (connector scoping, audit logs)
Best for	Customer-facing AI features	Lead outreach, support triage, metrics reports

The decision in one sentence

If you’re a software company building an AI feature into your product, start with Agent Kit. If you’re any kind of team wanting to automate your own internal workflows, start with a managed agent product. The two can coexist in the same company, and many do.

What we do with each

We implement AI in business and pick the right model per task (ChatGPT, Claude, GigaChat, YandexGPT). Our focus is internal team automation: done for you, in a week. If you need an Agent Kit build (a customer-facing AI feature inside your own product), that is a different, engineering-heavy skill set, and we’ll point you to specialists we trust rather than pretend otherwise.

Questions

Need an internal AI agent shipped?

20-min intro call. Done for you, in a week, priced on brief. If it turns out you actually need Agent Kit instead, we’ll tell you on the call.

Book a 20-min intro call Or send a note

The AI agents we ship →