Skip to content
Independent guide · Dots Bot

Developer & API Guides

Choose the right OpenAI building block for an agent you are developing.

Updated 4 October 2026 · Independent of OpenAI

Developer & API Guides — Build with the right agent tools
Visual guide to Developer & API Guides

Agents API, Agents SDK, and Responses API offer different levels of managed orchestration. These developer tools are distinct from the consumer dot experience.

Choose the right OpenAI building block for an agent you are developing. Treat the first output as a draft to interrogate. Look for its source trail, what it left out, and the decision it has handed back to you.

Editorial route

Choose who owns the agent loop

The developer paths here differ in state management, tool execution, environment, and observability. A direct model request is useful for one step; a managed session can support a continuing task; a reusable skill packages instructions rather than permissions. The Decisions API announcement is a separate preview topic. Start with the intended application behavior and threat model, then use the relevant official documentation to implement the smallest route that fits.

Choose a route

Guides in this section

Pick the specific page that matches the work in front of you. Each guide has a flat URL, a concrete review point, and links back to this overview. If the request spans several guides, finish one bounded piece before adding the next.

Define the task

How to use this section

The essential inputs are application goal, runtime control, tools, budget. Describe what is missing explicitly; a blank field is safer than a plausible value borrowed from another context.

A useful deliverable

A documented architecture choice. A compact output with evidence is easier to evaluate than a long answer that hides its assumptions.

Worked situation

How this plays out in practice

A team can prototype a managed agent session, a custom SDK orchestration, or a direct model request. Those choices affect where state, tools, and approvals live. A response that misses that detail may still look fluent. Make the first pass identify the dependency, and use a second pass only after resolving it.

Begin with the smallest real workflow, then map environment, tool access, session duration, logs, and human review before selecting an API path. Mark the point at which a draft became a human-approved decision. That boundary matters when the task returns.

A practical sequence

Run a bounded first pass

1. Establish the starting point

Name the period you are studying and assemble application goal, runtime control, tools, budget. Check which item is current before accepting a convenient older copy.

2. Request a reviewable draft

Begin with the runtime choice: OpenAI-managed Codex harness, your application-managed agent loop, or direct model requests. Then choose tools, environments, observability, and review points. Have uncertain items collected below the result instead of being silently filled in.

3. Decide the next action

Before the next action, apply this boundary: do not present api features as consumer dot controls. Read the proposed target and effect, then decide whether to authorize it.

Editorial judgment

Where a plausible answer can fail

A product announcement is not an endpoint specification. Use the current developer reference for implementation details. Ask what evidence would falsify the proposed answer. That question often exposes the missing piece.

Use current API docs for schemas and prices. Do not copy a consumer Dots instruction into an API request and expect the same product behavior. If two sources disagree, preserve both with their dates and ask who owns the final interpretation.

Where to go next

Connect this result to the right follow-up

If you are new to the product, Get Started & Trust helps establish the account and permission basics before you use this section as a working plan. Related work should inherit only the context it needs. A fresh guide gives you a chance to reset the audience and action boundary.

For a closely related outcome, open OpenAI Agents API Explained. That guide focuses on the Agents API exposes an OpenAI-managed Codex harness to applications. It may use some of the same context, but you should confirm the source and the decision point again. If your next question is about a different angle, Agents API vs Agents SDK vs Responses API is a better fit: compare three OpenAI paths by who manages the agent loop. A new deliverable should have its own date, owner, and acceptance check.

Before you reuse this result elsewhere, look at Get Started & Trust for the wider product or review boundary. Take forward the verified facts and leave behind any assumption that was specific to this example. If the source has changed since the draft was made, update the relevant claim first. The purpose of the link is to make the next step more precise, not to imply that every connected guide is needed for every reader.

Adapt this

Build your own task brief

Edit the bracketed information and any sentence that does not fit your account. This is a suggested workflow, not a promise that your dot has access to a particular app or feature.

Your review

Check before moving on

Tick what you have actually checked. Your progress stays on this device for this page and can be cleared at any time.

0 of 4 checked

Questions

Frequently asked questions

Which guide should I open first?

Open the child guide closest to the result you need: a documented architecture choice. Start with its inputs and review gate, then come back to this overview for adjacent tasks.

What information matters most for developer & api guides?

Start with application goal, runtime control, tools, budget. Agents API, Agents SDK, and Responses API offer different levels of managed orchestration. These developer tools are distinct from the consumer dot experience. Give the exact source or account where a decision depends on it.

How do I combine two related workflows?

Define separate results and approval points. Use current API docs for schemas and prices. Do not copy a consumer Dots instruction into an API request and expect the same product behavior.

What would a useful result look like here?

A documented architecture choice. It should make assumptions visible and leave a trail to the source, especially for dates, prices, permissions, or changing product details.

What is the common failure in this situation?

A product announcement is not an endpoint specification. Use the current developer reference for implementation details. Review that issue in the actual output instead of treating a polished presentation as proof.

Where can I check the current details?

Follow the official source linked below and inspect your account where access matters. This page was revised in October 2026; availability, terms, and product behavior can change.

Source for current product details: official OpenAI documentation or announcement ↗. The worked situations and briefs on this independent site are editorial examples. Confirm current access and details in your account. Editorial policy.