Agent Builder

Enterprise

OpenAI

Product overview

Name of Agent: Agent Builder
Short description of agent: "Agent Builder is a visual canvas for building multi-step agent workflows." (link, archived)
Date of release: 06/10/2025 (link, archived)
Advertised use: "complete set of tools for developers and enterprises to build, deploy, and optimize agents" (link, archived)
Monetisation/Usage price: API pricing, extra pricing for AgentKit storage (link, archived)
Who is using it?: developers through API and enterprise developers (link, archived))
Category: Enterprise

Company & accountability

Developer: OpenAI
Name of legal entity: OpenAI, L.L.C. (link, archived)
Place of legal incorporation: Delaware
For profit company?: Yes
Parent company?: For-profit LLC falls within the OpenAI Group (PBC) which is controlled by OpenAI Foundation (26% vs Microsoft's 27%, rest going to staff)
Governance documents analysis: Terms and Policies (link, archived)(general to OpenAI, not product specific)
AI safety/trust framework: Preparedness Framework (link, archived)
Compliance with existing standards: unsure, likely same as ChatGPT

Technical capabilities & system architecture

Model specifications: OpenAI models (link, archived)
Observation space: Text and images, user specified data from third party apps (link, archived)
Action space: Number of tools available, details (here, archived), including internet access and sandboxed code execution.
Memory architecture: Memory tool can be implemented through the MCP (link, archived)
User interface and interaction design: Node-based canvas interface
User roles: Designer (user designs agentic workflows), Operator, Executor, Examiner
Component accessibility: Closed source

Autonomy & control

Autonomy level and planning depth: L1: user has full control over how to design the agent and is directly manipulating the canvas. However, agents designed by the tool seem to be around L4 (main user interaction is approve/disapprove)
User approval requirements for different decision types: Special Human approval node (link, archived); safety guidance recommends keeping tool approvals on (especially for MCP tools) so users can review/confirm operations (link, archived)
Execution monitoring, traces, and transparency: Visible CoT and action trace documentation, also documents underlying decisions in JSON representation (link, archived)
Emergency stop and shut down mechanisms and user control: User can pause/stop the agent at any time
Usage monitoring and statistics and patterns: Nothing specific, uses stats/dashboards available in the OpenAI Developer Platform

Ecosystem interaction

Identify to humans?: None
Identifies technically?: None found
Interoperability standards and integrations: - MCP support (link, archived)
Web conduct: None found

Safety, evaluation & impact

Technical guardrails and safety measures: AgentKit’s Agent Builder includes a Guardrails node that can be added to workflows to monitor for unwanted inputs/outputs such as PII, jailbreaks, hallucinations, and other misuse (link, archived)
Sandboxing and containment approaches: AgentKit’s Code Interpreter (“python tool”) runs Python in a sandboxed environment (link, archived)
What types of risks were evaluated?: None found
(Internal) safety evaluations and results: None found
Third-party testing, audits, and red-teaming: None found
Benchmark performance and demonstrated capabilities: None found
Bug bounty programmes and vulnerability disclosure: None found
Any known incidents?: None found
← Back to 2025 Index