There are over 70 coding brokers in the marketplace: Claude Code, Codex, Cursor, Copilot, Devin, and an extended tail of smaller entrants exhibiting up in engineers’ newsfeeds each week. Somebody runs a benchmark, posts the leaderboard, and by the tip of the week, half the org needs to strive no matter is in first place. The tempo behind that impulse is actual: these instruments are enhancing each few weeks.
The issue is what “attempting a brand new mannequin” really prices. Claude Code solely runs Anthropic’s fashions. Codex solely runs OpenAI’s. Choose one and also you’re locked into the mannequin supplier that comes with it.
If you wish to strive a unique mannequin, akin to one thing open-weight and cheaper for a given job, you’re not simply altering a setting. You’re switching the entire instrument. Chasing a mannequin, value level, or benchmark means adopting a brand new coding agent, and a brand new vendor.
In organizations, that’s the place the evaluation exhibits up. Each new coding agent presents the identical three questions:
- Does it prepare in your code by default?
- The place does inference bodily run?
- Is pricing per seat or per token?
The solutions change instrument to instrument, however the evaluation itself is sort of similar each time, and authorized, safety, and finance find yourself doing it from scratch every time.
To bridge this hole, DataRobot OpenCode delivers a state-of-the-art coding agent that helps mannequin selection throughout closed, open, and bring-your-own choices, all whereas sustaining rigorous governance.
Getting began solely takes a couple of minutes.
How DataRobot OpenCode works
DataRobot OpenCode runs on the DataRobot LLM Gateway, the identical ruled entry level already used for each mannequin on the platform, so the agent inherits that governance as a substitute of routing round it.
All of it sits behind one already-approved key. Switching from a closed frontier mannequin to an open-weight one for a cost-sensitive job is a config change, not a brand new vendor relationship.
DataRobot agent abilities, together with Agent Help, come preinstalled, so OpenCode is enterprise-ready on day one, not simply model-flexible. That’s the identical abilities layer DataRobot ships for Cursor, Claude Code, and Gemini, utilized right here to an agent that isn’t tied to any single vendor’s fashions.
Set up is three instructions from the DataRobot CLI. No separate API key to request, no mannequin configuration to write down by hand.
You’ll be able to set up and launch OpenCode in your terminal or run it from the DataRobot UI:
brew set up datarobot-oss/faucets/dr-cli
# installs the DataRobot CLI in your terminal; not wanted contained in the DataRobot UI
dr plugin set up opencode
# installs the OpenCode plugin
dr opencode
# launches OpenCode, already wired to the Gateway with DataRobot abilitiesChoose a mannequin from the LLM Gateway through:
/fashionsThen, launch the Agent Help ability to make use of pure language to design, take a look at, construct, and deploy production-grade brokers:
/datarobot-agent-assistAttempt it
Get began right now. Run the next instructions in your terminal:
brew set up datarobot-oss/faucets/dr-cli
dr plugin set up opencode
dr opencode
