OpenAI AgentKit review
★ 4.2/5Agent builder
OpenAI's visual, drag-and-drop toolkit for building, deploying, and optimizing AI agents without standing up your own orchestration layer. It bundles Agent Builder (a node-based canvas for multi-agent workflows with guardrails), ChatKit (embeddable chat UI), Connector Registry (managed data connectors), and Evals. There is no separate product fee: you build for free and pay only standard OpenAI API usage when your agents actually run.
Affiliate link · how we make money
Best for Teams already on the OpenAI API who want a visual agent canvas. OpenAI AgentKit starts Free to build; usage ~$0.20/1M input tokens (GPT-5.4-nano) on a Usage-based (pay-per-API-token plus built-in tool fees); no platform subscription. Agent Builder design time is free; you pay standard OpenAI API rates only when workflows run. model. There is no fee for AgentKit itself; cost = OpenAI API usage, e.g. a GPT-5.4 agent run is $2.50 per 1M input tokens and $15.00 per 1M output tokens, plus tool fees like Web Search at $10 per 1K calls and Code Interpreter at $0.03 per 20-minute 1 GB session.
Spec
The gotcha
Pricing is hard to predict because it is purely usage-based: multi-agent workflows fan out across many model and tool calls, so token spend compounds fast. ChatKit and File Search storage are free only up to 1 GB per account per month, then $0.10/GB-day. Agent Builder is still in beta, and Connector Registry is rolling out to API, ChatGPT Enterprise and Edu accounts.
Decode the real cost →Best for
- You want to visually build and ship multi-agent workflows on OpenAI models without writing orchestration code or paying a separate platform fee.
source: openai.com/pricing →
OpenAI AgentKit pricing tiers
Prices move. Cells flagged verify link to the live vendor page.
| tier | monthly | annual | included | unit |
|---|---|---|---|---|
| Build (Agent Builder design time) | $0 | n/a | Visual drag-and-drop canvas, preview runs, versioning, guardrails | flat |
| GPT-5.4 (typical agent model) | Usageverify | n/a | $2.50 / 1M input tokens, $15.00 / 1M output tokens | per 1M tokens |
| GPT-5.4-mini | Usageverify | n/a | $0.75 / 1M input tokens, $4.50 / 1M output tokens | per 1M tokens |
| GPT-5.4-nano | Usageverify | n/a | $0.20 / 1M input tokens, $1.25 / 1M output tokens | per 1M tokens |
| Built-in tools | Usageverify | n/a | Web Search $10 / 1K calls; File Search $2.50 / 1K calls + $0.10/GB-day storage (1 GB free); Code Interpreter $0.03 per 20-min 1 GB session | per use |
| ChatKit storage | Usageverify | n/a | $0.10/GB-day after 1 GB free per account/month (billing began Nov 1, 2025) | per GB-day |
// pools and per-unit rates are volatile · cells flagged verify link to the live vendor page
Compare OpenAI AgentKit with…
Questions
How much does OpenAI AgentKit cost?
OpenAI AgentKit starts Free to build; usage from ~$0.20/1M input tokens (GPT-5.4-nano). There is no fee for AgentKit itself; cost = OpenAI API usage, e.g. a GPT-5.4 agent run is $2.50 per 1M input tokens and $15.00 per 1M output tokens, plus tool fees like Web Search at $10 per 1K calls and Code Interpreter at $0.03 per 20-minute 1 GB session.
Can you self-host OpenAI AgentKit?
No, OpenAI AgentKit is cloud-only.
Does OpenAI AgentKit support MCP?
OpenAI AgentKit supports MCP as Client, so it works with assistants like Claude and ChatGPT.
What's the catch with OpenAI AgentKit's pricing?
Pricing is hard to predict because it is purely usage-based: multi-agent workflows fan out across many model and tool calls, so token spend compounds fast. ChatKit and File Search storage are free only up to 1 GB per account per month, then $0.10/GB-day. Agent Builder is still in beta, and Connector Registry is rolling out to API, ChatGPT Enterprise and Edu accounts.