Existing workflow owners
Teams with live Agent Builder canvases need the remaining window to inventory behavior, preserve test cases, and complete a migration.
Independent tool overview
Agent Builder at a glance
OpenAI Agent Builder is a visual multi-agent workflow canvas that remains available temporarily but is being retired; OpenAI says it will no longer be available after November 30, 2026.
Visit the official Agent Builder site ↗
- Developer
- OpenAI
- Launched
- October 6, 2025 in beta
- Current state
- Available during wind-down
- Shutdown
- Unavailable from November 30, 2026
- Recommended code migration
- OpenAI Agents SDK
- Separate license fee
- None; standard API usage charges apply
Overview
What Agent Builder is
OpenAI launched Agent Builder in beta as a drag-and-drop canvas for composing agent logic, connecting tools, adding guardrails, previewing runs, configuring evaluations, and versioning multi-agent workflows. It was part of AgentKit alongside ChatKit and Connector Registry and used standard OpenAI API models and tools rather than a separate proprietary runtime charge.
The product is now in wind-down. OpenAI added a June 3, 2026 notice stating that Agent Builder and its Evals product will be unavailable on the platform from November 30, 2026 onward. Do not begin a new production dependency on it. Existing users should document their workflow behavior, test cases, tools, prompts, permissions, and failure paths, then rebuild workflows that must continue in the OpenAI Agents SDK or, for natural-language team workflows, evaluate Workspace Agents in ChatGPT.
Use cases
Who Agent Builder is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
Migration planning
The visual graph can help engineers and domain experts identify agents, tools, handoffs, guardrails, and decision branches that code must reproduce.
Short-lived demonstrations
A disposable internal demo may still illustrate an old workflow, provided no one mistakes it for a durable production platform.
Historical architecture reference
Teams evaluating AgentKit-era projects can use the remaining interface and launch materials to understand how the original visual workflow model worked.
Capabilities
Core Agent Builder features
Visual workflow canvas
Agent logic was composed with connected nodes instead of being written entirely as orchestration code.
Multi-agent workflows
The canvas supported flows involving more than one configured agent and explicit transitions between workflow steps.
Tool connections
Builders could connect agent nodes to tools needed for retrieval, business logic, or other actions.
Custom guardrails
The original product supported guardrail configuration for checks such as identifying sensitive data or unwanted behavior.
Preview runs
Teams could exercise a workflow in the builder before using it in an application.
Inline evaluation setup
The launch materials described evaluation configuration alongside the workflow, although that platform product is also being retired.
Workflow versioning
Agent Builder tracked workflow versions to support iteration and deployment control.
Templates
Users could begin from a blank canvas or prebuilt workflow templates.
Process
How the Agent Builder workflow works
Step 1
Freeze new dependency growth
Avoid adding new production workflows or major features to a product with a confirmed shutdown date.
Step 2
Inventory every workflow
Record owners, users, models, prompts, nodes, tools, credentials, guardrails, versions, inputs, outputs, and downstream consumers.
Step 3
Capture behavioral tests
Save representative inputs, expected results, tool-call rules, approval points, failures, and traces without copying secrets into documentation.
Step 4
Choose the successor
Use the Agents SDK for workflows that must continue as code; assess Workspace Agents for team workflows better expressed through natural-language instructions.
Step 5
Rebuild behind a stable interface
Keep the application's external contract stable while recreating orchestration, tools, handoffs, guardrails, state, and tracing in the replacement.
Step 6
Run parallel verification and retire
Compare old and new behavior on the captured test set, move traffic gradually, verify observability and rollback, then decommission Agent Builder dependencies before November 30.
Cost
Agent Builder pricing and free plan
Agent Builder has no separate license price. During its remaining availability, model tokens and API tools are billed at OpenAI's standard API rates. Migration to the open-source Agents SDK also has no SDK license fee, but API, tool, storage, and infrastructure charges remain.
Agent Builder beta
No separate platform fee
The visual builder is included with standard OpenAI API usage pricing during the wind-down period.
- Only temporarily available
- No longer available from November 30, 2026
- Model tokens and tools are billed separately
OpenAI API usage
Usage based
Actual cost depends on the selected models, token volume, built-in tools, storage, and external systems used by the workflow.
- Model rates vary
- Web search, file search, code execution, or other tools can add charges
- Failed and repeated agent runs still consume resources
OpenAI Agents SDK
No SDK license fee
The recommended code-based migration path is available for Python and TypeScript.
- OpenAI API usage remains billable
- Your hosting, databases, queues, and monitoring add operational cost
- Migration engineering is not included
Pricing checked . Check current pricing at the source ↗
Assessment
Agent Builder strengths and limitations
Where it stands out
- Visual graph made multi-step agent logic easier for technical and nontechnical stakeholders to inspect together
- Combined agents, tools, guardrails, previews, evaluations, and versions in one interface
- Templates offered a faster starting point than a blank orchestration implementation
- Preview runs helped expose workflow behavior before application integration
- Versioning gave teams a defined artifact to compare and discuss
- OpenAI has provided a clear shutdown date and named migration paths
What to consider
- OpenAI is winding down Agent Builder and it will be unavailable from November 30, 2026
- It is not a suitable foundation for new production work because continued availability is time-limited
- The associated platform Evals product is also being wound down, so evaluation dependencies need separate migration planning
- A visual graph does not remove the need for authentication, authorization, retries, idempotency, state management, queues, and failure recovery
- Agent outputs and tool choices remain probabilistic and require representative evaluations
- Tool-connected workflows can be exposed to prompt injection, excessive permissions, and unsafe side effects
- Guardrails reduce particular risks but do not prove a workflow is safe, correct, compliant, or secure
- Standard API charges can grow quickly when workflows branch, retry, use expensive models, or call paid tools
- Migration to code requires engineers to reproduce behavior that may have been implicit in nodes, defaults, or platform state
- Tracing may contain sensitive prompts, tool inputs, or outputs unless collection and retention are configured carefully
- Visual versioning is not a substitute for source control, code review, automated tests, release controls, and rollback
- Teams must finish migration and runtime verification before the shutdown rather than treating November 30 as the day to begin
Compare
Agent Builder alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Agents
Workspace Agents
OpenAI explicitly recommends Workspace Agents for team workflows that are better expressed through natural-language prompting than orchestration code.
Explore Workspace Agents →Business Operations
n8n
n8n provides a general visual automation canvas with self-hosting options and a broader ecosystem of deterministic workflow integrations.
Explore n8n →Agents
Agent Workflows
ElevenLabs Agent Workflows is a visual builder focused on conversational and voice-agent flows.
Explore Agent Workflows →Coding
AgentOps
AgentOps focuses on tracing, testing, and debugging agent systems when observability is the most urgent migration need.
Explore AgentOps →Questions
Agent Builder FAQs
Is OpenAI Agent Builder shutting down?
Yes. OpenAI says Agent Builder is being wound down and will no longer be available on the OpenAI platform from November 30, 2026 onward.
Is Agent Builder still available today?
As of August 30, 2026, OpenAI describes it as being in wind-down rather than already unavailable. Existing users should use the remaining time for migration, not expansion.
What should replace Agent Builder?
OpenAI recommends the Agents SDK for workflows that need to continue as code and Workspace Agents in ChatGPT for use cases better suited to natural-language prompting.
What did Agent Builder do?
It provided a visual canvas for composing multi-agent logic with tools and guardrails, previewing runs, configuring evaluations, starting from templates, and versioning workflows.
How much does Agent Builder cost?
It has no separate platform charge in the launch pricing. Model tokens and tools are billed at standard OpenAI API rates, and that usage continues to matter during migration testing.
Can an Agent Builder workflow be left running after November 30, 2026?
Do not plan on it. OpenAI says the product will no longer be available from that date. A workflow that must continue needs a verified replacement before the shutdown.
Does migrating to the Agents SDK eliminate OpenAI API charges?
No. The SDK itself has no license fee, but OpenAI model and tool usage remains billable, and your own hosting, storage, monitoring, and engineering have costs.
What should be captured before migration?
Inventory prompts, models, nodes, tools, schemas, permissions, guardrails, versions, state, approval gates, error paths, representative inputs, expected outcomes, traces, owners, and downstream dependencies.
Bottom line
Our Agent Builder verdict
Agent Builder was a useful visual prototyping and collaboration layer, but its confirmed shutdown dominates any present-day evaluation. Existing users should treat it as a migration source, not a platform to grow. Start rebuilding critical workflows in the Agents SDK now, verify them in parallel, and leave time for a controlled cutover before November 30, 2026.
Visit Agent Builder website ↗Similar Agents Tools
View all Agents AI tools →
Anything
Anything - AI agent that ships mobile and web apps without coding
Petri
Petri - Anthropic's open-source agentic system for evaluating LLM safety
Action Agent
Action Agent - Writer's enterprise-ready autonomous AI that works on your behalf
Gemini Computer Use
Gemini 2.5 Computer Use - A specialized model to power agents that can interact with user interfaces
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.