Oracle opens Fusion to coding agents, and Codex and Claude Code walk in
The ERP giant is inviting other companies' agents to build production apps inside its system of record. Agents building applications for agents, under Fusion's governance.
- Oracle announced an AI-native builder experience for its AI Agent Studio on July 14, extending Fusion Agentic Application development from no-code users to pro-code developers and to AI coding agents themselves.
- Supported coding agents explicitly include OpenAI's Codex and Anthropic's Claude Code, working through VS Code, standard CLIs, and Git workflows.
- A new AI Studio Skill gives coding agents Fusion-specific context: project structure, APIs, templates, validation, packaging, and deployment.
Oracle just formalized a milestone that sounds like science fiction when said plainly: other companies' AI agents are now invited to build production applications inside its ERP. The company announced a new AI-native builder experience for Oracle AI Agent Studio on July 14, extending Fusion Agentic Application development beyond the no-code builder it shipped in March to pro-code developers and, notably, to AI coding agents, with OpenAI's Codex and Anthropic's Claude Code named as supported.
A skill that teaches agents to speak Fusion
The centerpiece is the AI Studio Skill, a CLI that loads Fusion-specific context into a compatible coding agent: project structure, APIs, templates, validation, packaging, and deployment workflows. The intended loop, as SiliconANGLE describes it, is that a developer describes a business outcome, hands the skill to an agent, and the agent generates, validates, and tests the Fusion artifacts before deployment. "We're now going to go after what's called pro-code people," Oracle SVP Natalia Rachelson told the outlet. A public GitHub repository with templates and reference architectures rounds out the developer surface, and everything the agents build inherits Fusion's security, governance, and auditability.
The system of record becomes the runtime
The strategic read is that Oracle wants Fusion to be both the system of record and the platform where autonomous enterprise apps run, with agents operating on live transactional data under its governance instead of somebody else's. It is a different answer to the question the whole industry is circling, the one the superagent standards fight is really about: who owns the surface where agents do enterprise work. And it compounds the pattern from Claude Code's feature sprints to OpenAI's in-API orchestration: the coding agent is becoming the universal contractor, and now the biggest back offices on earth have a door for it.