Launch Directory/AI Agents/Openai Agents Api Public Beta (by OpenAIDevs)
Openai Agents Api Public Beta (by OpenAIDevs)

Openai Agents Api Public Beta (by OpenAIDevs) · Launch Video Breakdown: Hook, Pacing & Motion Design

Go from idea to a working agent faster with the Agents API. Build and run cloud agents with the Codex harness, fully managed by OpenAI. We handle orchestration, long-running sessions, and context ma

0:00 · The Hook · Challenges of Production Agents
0:00 / 0:00
AI AgentsPublic LaunchSeptember 10, 2026

Problem Agitation

7.8

Cuts / min

2:18

Runtime

1.9M

Views

4,236

Likes

374

Reposts

225

Replies

Scene-by-scene timeline & spoken transcript

  1. The Hook

    Challenges of Production Agents

    “Getting a long-running agent into production takes a lot of work, even with a capable model. You need to connect tools, track progress, manage context, and secure and maintain the infrastructure around it.”

    On screen
    Charlie Guo Developer Experience Engineer, OpenAI
    Camera
    Static wide shot centered on presenter at desk
    Motion
    Lower-third typography fade-in
  2. Product Reveal

    Agents API Reveal

    “Today, we're launching the Agents API to handle that infrastructure for you.”

    On screen
    Agents API
    Camera
    Medium close-up push-in
    Motion
    Subtle text tracking and clean cut
  3. Feature Teaser

    Codex Harness Architecture

    “The Agents API brings a hosted version of the Codex harness to your applications. OpenAI handles orchestration, sessions, and context management so that you can stay focused on building.”

    On screen
    Your app Harness Orchestration Sessions Context management Sandbox
    Camera
    Fixed graphic frame
    Motion
    Clean line-draw diagram animation with progressive text reveal
  4. Feature Teaser

    Incident Investigation Demo & MCP Tools

    “Let's look at an example. Suppose we want to build an agent that helps investigate incidents in our production stack. It would need access to observability data and recent code changes, along with our team's instructions for how to handle an outage. We can connect all of our necessary tools through MCPs and give the agent our investigation runbook via a skill.”

    On screen
    session = await client.sessions.create( capability_directories = [ "/plugins/datadog", # MCP tools "/plugins/github", # Skills + MCP tools "/skills", # Investigation runbook ] # Investigate a checkout incident
    Camera
    Alternating presenter split-screen and terminal close-ups
    Motion
    Typewriter terminal streaming and code window overlays
  5. Feature Teaser

    Flexible Execution Environments

    “You control the agent's execution environment and the tools it has access to. That includes connecting a sandbox, whether it's through OpenAI, a third-party provider, or using your own infrastructure.”

    On screen
    # OpenAI-hosted sandbox # Third-party sandbox # Your infrastructure
    Camera
    Medium shot with dynamic left-third code panel
    Motion
    Syntax-highlighted code block transitions
  6. Feature Teaser

    Programmatic Tool Calling

    “In a situation like this, we'd also be working through a ton of logs, likely more than we could fit in the model's context window. With programmatic tool calling, the agent can process those logs and filter the results in code, meaning fewer tokens spent passing around raw data and more spent on the information the agent needs.”

    On screen
    Programmatic tool calling With direct tool calls 170,126 Tool-result tokens With PTC 11 Tool-result tokens
    Camera
    Full-screen graphic display with terminal cutaway
    Motion
    Side-by-side data comparison and metric counter highlight
  7. Feature Teaser

    Multi-Agent Orchestration

    “For the largest tasks, independent work can be delegated via multi-agent orchestration. In this case, we might want one sub-agent to examine recent changes, another to check telemetry, and then have the lead agent unify their findings.”

    On screen
    spawn_agent_call subagent_01 subagent_02 Code changes: the upstream timeout fell from 2,000 ms to 200 ms. Telemetry: 42 errors begin after deployment; upstream p95 is 380 ms.
    Camera
    Terminal output stream into medium speaking shot
    Motion
    JSON event stream telemetry display
  8. Feature Teaser

    Context Compaction & Final Report

    “But even in a single agent session, long-running context windows still work really well with compaction. It gives the model a summary of all prior work completed so the agent can continue its investigation. Once the work is complete, the findings should be served as a report that the on-call team can review: a likely root cause, the supporting evidence, and suggested next steps bundled into a single shareable file.”

    On screen
    Checkout incident Deployment 1842 Assessment 42 checkout errors found 200 ms new timeout 380 ms upstream p95
    Camera
    Direct terminal view scrolling down to a document PDF view
    Motion
    Screencast smooth vertical scroll across PDF deliverable
  9. Call to Action

    Ever-Improving Harness Outro

    “And there we have it. With the Agents API, we were able to drive an entire workflow without building or maintaining any of our own agent infrastructure. Infrastructure that will continue to get better alongside new models and new capabilities. We're excited to bring you an ever-improving harness behind one API. Happy building.”

    On screen
    OpenAI
    Camera
    Medium close-up to wide shot, transitioning to white background
    Motion
    Geometric circle drop transition into minimal logo animation

Related AI Agents Product Launches

Explore all AI Agents launches →
OpenAI
@OpenAI
AI Agents
@OpenAI
OpenAI@OpenAI
Gopuff
@gopuff
AI Agents
@gopuff
Gopuff@gopuff
Elon Musk
@elonmusk
AI Agents
@elonmusk
Elon Musk@elonmusk