Is OpenAI AgentKit free? What it actually costs in 2026
AgentKit has no plan of its own: an agent pays the model's per-token rates plus per-call tool fees — web search $10 per 1,000 calls, file search $2.50 per 1,000 calls and $0.10 per GB a day of storage after 1 GB, hosted containers from $0.03 per 20-minute session.
The pricing page's Tools table is the AgentKit bill. Tokens for built-in tools are charged at the chosen model's rate; ChatKit file and image uploads get 1 GB free per account per month, then $0.10 per GB-day. The Agents, Responses and Realtime APIs carry no separate fee.
What the free tier actually includes
- No free tier; tokens at model rates on every step
- File search storage: 1 GB free, then $0.10 per GB per day
- ChatKit uploads: 1 GB free per account per month, then $0.10 per GB-day
Where the wall is
Tool calls at volume. An agent that searches the web ten times per task pays $0.10 in search fees before a single output token; file search adds $2.50 per thousand calls plus daily storage. The meter is per action, and agents take many.
After that: Web search $10 per 1,000 calls ($25 on the non-reasoning preview); file search $2.50 per 1,000 calls + $0.10 per GB-day; Hosted Shell / Code Interpreter containers $0.03 (1 GB) to $1.92 (64 GB) per 20-minute session; model tokens at the model's standard rates.
Can you self-host OpenAI AgentKit? No — cloud only
AgentKit runs on OpenAI's platform. LangGraph, CrewAI and OpenClaw are the open agent frameworks compared here; they run wherever your code runs and call any model.
What it actually takes to run one yourself
Real RAM against the documented minimum, monthly cost, setup time, and the one thing that catches people out.
When paying is still the right call
Through November 2026, if what you have already works and rewriting it now costs more than running it out — there is no benefit in a panic migration for a system that will be replaced anyway. Stay too if you are prototyping this week and expect the result to be thrown away: managed tooling is genuinely faster for that, and a throwaway prototype has no migration cost by definition. And for the rest of OpenAI's API — models, ChatKit, the Responses API — nothing here is a reason to leave. Only the agent-building layer is closing.
What you are locked into
The prompts, the tool definitions and the model calls are portable — those are yours and they move. What does not move is the orchestration: a graph built in Agent Builder is a description held in OpenAI's product, not a file in your repository, and rebuilding it elsewhere means rebuilding the logic rather than exporting it. Evals histories are similar. This is the specific shape of visual-builder lock-in: the easier a platform makes it to avoid writing the wiring down, the less of the wiring you own when it leaves.
If you would rather not pay: LangGraph
LangGraph
Agents as an explicit state graph — the closest thing to what you are leaving.
Free and open source. LangGraph Platform is an optional paid hosted tier.
See all 4 OpenAI AgentKit alternatives comparedCompare the free options head-to-head
Read from OpenAI's own pricing and documentation on 2026-09-26 — 33 days ago. Vendors change prices without notice; confirm before you commit.