Your Data Team should be able to take a business question, figure out what work it requires, and bring in the right specialists. You should not have to manage the handoffs.

This release makes Dreambase better at that job. Our primary agent now coordinates more of the work across your Data Team, both in the app and through MCP. Metrics become first-class workspace items you can run the business on. API and MCP connections gain a deeper Semantic Layer. Together, these changes give your Data Team more context to work with and more responsibility for maintaining the data you rely on.

Smarter Orchestration Across Your Data Team

Dreambase’s primary agent has become a more capable orchestrator. It interprets your request and routes the work to the appropriate subagent, whether you are preparing a dataset, investigating a number, or creating an analytical view. You work through one conversation while Dreambase coordinates the specialists behind it.

That coordination now carries more work through to completion. In workspace chat, the agent can plan an analytical view, show you what it intends to build, and open a live preview beside the conversation. When a task pauses or fails, it explains why and offers a way forward in the thread.

(The data team hard at work coordinating KPIs & Metrics)

Setup follows the same direction. Dreambase starts with your business context, connects your sources, and shows an initial analysis using your own numbers. It identifies measurable KPIs, prepares the analytical work behind them, and recommends additional sources where the available data falls short. Questions and decisions appear on cards so the next step is clear.

Run the Business on Metrics

Our customers want to run their businesses on Metrics. Those Metrics need to be accessible wherever a decision happens.

Metrics are now first-class items in Dreambase. They belong to the workspace rather than living under an individual dashboard, with their latest calculated values available across metric lists, detail views, and dataset pages. A dashboard can display a Metric, but it no longer defines where that Metric lives.

Scheduled dataset refreshes now record each associated Metric’s value, building its own history over time. Durable datasets without a dashboard also receive daily refresh schedules, so the analytical work can continue independently of how it is presented. History recording starts at deployment, with access in the interface and chat rolling out separately.

(Preview of the upcoming Dreambase mobile app)

This also sets up what comes next: the Dreambase iOS app, built on Expo. We are building it for business stakeholders who want their KPIs at a glance, on their phone, wherever they are. More on that soon.

A Semantic Layer for APIs and MCP

Connecting an API or MCP server should give your Data Team an understanding of the data behind it, not just a list of endpoints and tools.

Dreambase now scans selected read endpoints and MCP tools, samples their results, profiles the returned data, and builds a semantic model of the entities and metrics it finds. This extends the schema-discovery approach we use for Supabase Postgres to sources that do not expose a conventional database schema.

Your Data Team gets more context about what each source contains and how its data can support the analysis. GitHub connections can also inspect schema files in recently pushed repositories. You can review the resulting Documentation and Semantic Layer in the connection’s Context Layer.

Ramp joins the connector catalog through its hosted MCP and OAuth with read-only scopes. We have also restored Polar installs and improved guided connector setup, including clearer Supabase permissions and next steps.

More Capacity for Analytical Work

Dataset reads now run in bounded DuckDB sessions over signed URLs, rather than downloading each dataset in full before execution. Concurrent work queues within defined memory and execution limits, and SQL dataset writes stream to temporary storage instead of accumulating in application memory.

For Supabase, workspaces can be configured with a dedicated read-only Postgres role for a much more optimal Postgres native connection that reads then run through Supavisor in read-only transactions, with timeouts, credential health checks, and smart fallbacks.

The Data Catalog That Builds Itself

As your Dreambase Data Team discovers sources, profiles their data, prepares datasets, and establishes Metrics, that work becomes part of a growing Data Catalog. Definitions, semantic models, and analytical context remain available for the next request.

The catalog becomes more useful as the team works. A new connection contributes an understanding of its data. A reusable dataset supports future investigations. A Metric accumulates history as scheduled refreshes run. Your people and their agents can work from that shared foundation rather than rebuilding it in each conversation.

That is the direction of Dreambase: a Data Team that does the analytical work and builds the catalog your business runs on as it goes.

Capabilities are rolling out by workspace. Some chat, connection, setup, and metric-history features may not appear everywhere immediately. The Dreambase iOS app is upcoming and is not part of this release.