DataRobot has introduced DataRobot OpenCode, a coding-agent integration designed to separate the developer interface from the model provider behind it. The release lets customers run OpenCode with closed models, open-weight models or models they bring to DataRobot, while routing requests through the company's existing LLM Gateway.
The company is positioning the integration as an answer to a growing enterprise problem: coding agents are often tied to one model family, so testing a different model can require adopting another tool, completing another security review and negotiating another pricing model. DataRobot says OpenCode makes the model a configuration choice inside one governed environment.
One gateway for multiple models
DataRobot OpenCode runs behind the same gateway that customers use for other models on the DataRobot platform. According to the announcement, developers can switch models through OpenCode's model menu without establishing a separate connection for each provider. That can make it easier to compare capability, latency and cost for different development tasks.
The approach does not make every model identical. Models can differ substantially in context limits, tool use, code quality, data handling and price. It instead gives organisations a common control point from which they can make and govern those choices. DataRobot says the gateway keeps access behind an already approved key, reducing the number of credentials and vendor-specific paths that teams need to manage.
DataRobot skills are included
The integration includes DataRobot agent skills and its Agent Assist workflow. Agent Assist is intended to help users design, test, build and deploy agents using natural-language instructions. The same skill layer is available for other supported coding environments, so organisations can reuse platform-specific guidance across tools.
DataRobot says OpenCode can run in a terminal or inside the DataRobot user interface. Terminal installation uses the DataRobot command-line interface and an OpenCode plug-in. After launch, developers can choose a model exposed by the LLM Gateway and invoke the Agent Assist skill when they need the platform's agent-building workflow.
Governance is the central proposition
For enterprises, a coding agent can read source code, call tools and produce changes that reach production systems. DataRobot's argument is that model choice should not bypass the controls applied to the rest of an organisation's AI use. Routing the agent through the LLM Gateway is intended to preserve the platform's existing governance rather than create a parallel path for each new coding tool.
Customers should still assess the behaviour of each model and agent configuration. A common gateway does not remove the need to review data retention, model deployment location, permissions, tool access, logging and generated changes. It also does not guarantee that a model will perform equally well across languages, repositories or security-sensitive tasks.
Availability and next steps
DataRobot says the OpenCode integration is available now and provides command-line installation instructions in its announcement. Existing customers will need access to the relevant DataRobot CLI, gateway models and skills. New users are directed to a trial, but the announcement does not provide a separate price for the OpenCode integration.
The release reflects a broader shift in coding agents from standalone assistants towards governed developer surfaces that can work with multiple models. For buyers, its value will depend on the models available in their DataRobot environment, the controls exposed by the gateway and whether centralising those controls materially reduces the work of evaluating and operating coding agents.