Today, Bobsled is launching the Bobsled Analytics MCP. It's a single endpoint that lets data teams make analytics available on-demand to agents without giving up control of how their data is accessed, interpreted, or used.

When you grant an agent access to Bobsled via MCP, it isn't talking to your APIs. It's talking to an Analytics Agent that already understands your semantic model, your data relationships, and your business context. Instead of exposing raw tables or fixed endpoints, teams building on Bobsled let downstream agents interact directly with the same tuned analytics agent that powers their existing agentic analytics experience.Today, we are thrilled to release the Bobsled MCP. It's a single endpoint that lets data teams make analytics available on-demand to agents without giving up control of how their data is accessed, interpreted, or used.

• What it is: A simple, secure way to give internal or external agents access to your data.
• Why now: Agents are coming for your data. Stakeholders love to chat with your data. But they love it even more when their agents do it for them.
• Why we're excited: When you grant an agent access to data products in Bobsled via MCP, it isn't talking to your APIs. It's talking to a Data Agent that already understands your semantic model, your data relationships, and your business context.

Get access now or keep reading for more details👇

Let agents chat with your data

In the past six months, nearly every data team has shipped a "chat with your data" capability. At Bobsled, we power those experiences for some of the most advanced data teams in the world.

But stakeholders are getting more sophisticated. They no longer want to just ask questions themselves. They want to build agents that can ask, decide, and act on data autonomously.

Most data teams have responded one of two ways.

They build one-off integrations for every new agent, or they wrap existing APIs in an MCP. Both approaches throw away the hardest-won asset the data team has: the context that makes the data useful. The endpoint returns rows, and the agent on the other side is left to guess what they mean.

When the consumer is an LLM, that guessing doesn't surface as a clarifying question. It gets silently resolved into a wrong answer. Multiply that across thousands of agent calls and you've recreated the data breadline that agentic analytics was supposed to eliminate, only with worse failure modes.

Improving agent performance with better data

The Bobsled Analytics MCP puts an intelligent agent on the other side of the connection, not just write SQL.

That agent handles ambiguity, generates and executes SQL dynamically, synthesizes results across tables, and returns answers in a language agents understand. The domain expertise your team has spent years encoding (how to navigate the model, what each field means, how to turn raw data into a real answer) stays embedded in the agent your customers' agents talk to. It isn't outsourced to theirs.

In practice, that means:

• Deeper analysis. Because the agent understands your full semantic model, it can answer questions that would never resolve through a fixed API endpoint: cross-table synthesis, trend analysis, comparative queries, and more.
• Richer responses. Anything the data agent can produce comes back over MCP. That includes natural-language explanations, structured results, and follow-up context, not just rows.
• Better accuracy. The agent knows your data the way your best analyst does. Downstream agents get answers that are precise, interpreted correctly, and grounded in the context your team has spent years building.

Governance stays where it belongs: with you. Access is scoped per data product, the agent only operates within the semantic model you've published, and every interaction is observable end-to-end.

Expanding access without hurting quality

Inside the next 12 to 24 months, agents will become the largest consumers of enterprise data. For data teams, the job is to support this rush without giving up the role of steward, the one that makes sure data is interpreted and used correctly.

The Bobsled Analytics MCP is how you do both. A single endpoint that turns the intelligence layer your team has already built into something every agent in your stack can reach.

Get started

Setup takes a few minutes. The Bobsled Analytics MCP works with Claude Code, Claude Desktop, Cursor, Windsurf, and any other MCP-compatible client. Reach out today and get started tomorrow.