OpenAI Launches Agents API in Public Beta, Opening the Codex Agent Harness to Developers
OpenAI's new Agents API, in public beta, turns the managed harness behind Codex into infrastructure any developer can build custom AI agents on, drawing mixed analyst reaction.
Overview
OpenAI on Wednesday introduced a new Agents API that brings the agent harness and infrastructure behind Codex to outside developers, according to InfoWorld. The API, now in public beta, lets developers build cloud-based agents that run for hours, execute code, and process files, according to The Decoder.
What We Know
The Agents API is a managed service, with OpenAI hosting and maintaining the underlying harness and infrastructure, according to InfoWorld. Previously, developers building a custom agent typically had to assemble an agent runtime, context and session management, tool and external data connections, and execution environments on their own, InfoWorld reports. With the new API, a developer can design a custom agent in a single API call after specifying the task, model, tools, and environment.
The service runs on the same infrastructure that powers Codex and ChatGPT and builds on the open-source Codex harness, according to The Decoder, which also reports that the API supports the Model Context Protocol, custom functions, and built-in tools such as web search. Key features include automatic context management, parallel tool use, and the ability to delegate tasks to sub-agents, per The Decoder.
For execution environments, developers can run agents in an OpenAI-managed sandbox, on their own infrastructure, or through supported sandbox providers including Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel, InfoWorld reports, giving enterprises the option to choose between fully managed environments and deployments within their own VPCs. On pricing, The Decoder reports there are no extra fees beyond standard token usage.
Industry Reaction
The Agents API “significantly reduces” engineering work, helping developers spend more time building the actual business application instead of the agent infrastructure, said Pareekh Jain, principal analyst at Pareekh Consulting, according to InfoWorld. “The main advantage with the Agents API is fewer moving parts. A long-running agent built by hand needs a job queue, a state database, a sandbox fleet, a compaction routine and a retry policy, and someone has to own and page for each of them,” said Amit Kumar Jena, AI development head at IT consulting firm Kanerika, per InfoWorld. Phil Fersht, CEO of HFS Research, added that “enterprises may need fewer engineers to build the infrastructure around each agent, which should reduce development time.”
That convenience carries tradeoffs, the analysts told InfoWorld. “Lock-in is the biggest concern. If OpenAI provides the model, context management, tools, orchestration, and execution environment, moving to another platform becomes harder,” Jain said. Jena separately pointed out that the new API “doesn’t support Zero Data Retention” even when enterprises use their own sandbox, an issue Fersht said could limit adoption “especially in industries in the regulated sectors, such as healthcare and BFSI.” Jain said startups, SaaS companies, and enterprises already using OpenAI are more likely to adopt the API readily, while “enterprises pursuing multi-model strategies may also prefer their own independent harness or are likely to take a hybrid approach.”
Competitive Landscape
Jena described the market as “a crowded category,” noting that Anthropic’s Claude Managed Agents has been in public beta since April — a service previously reported by The Machine Herald — and that Amazon Bedrock AgentCore’s managed harness became generally available in June, doing “the same category of work, meaning the loop, tool execution, context, state, and recovery, while letting you use any model and switch providers mid-session without losing context,” according to InfoWorld. Jena also named Microsoft’s Foundry Agent Service and LangGraph as additional competing offerings.
What We Don’t Know
Neither source details enterprise-tier pricing beyond the token-usage-only billing model, nor whether or when OpenAI might add Zero Data Retention support for the Agents API. How adoption plays out against Anthropic’s, Amazon’s, and Microsoft’s competing managed-agent services also remains to be seen.