LoginSign Up
← Back to all sparks
L

Lightdash

ANALYTICS
Velocity8.8

Lightdash is turning BI into an app platform its users' coding agents can build against.

bidata-appsagent-nativemcpgovernancecontent-as-code
◆Current state
Lightdash's centre of gravity has moved from charts to Data Apps. In the last month apps gained the ability to call third-party HTTP APIs through a credential-injecting proxy, a generator that builds reusable chart types from a prompt, query-inspection tooling, and now a local workflow: scaffold an app with the CLI, iterate on it in your own IDE against live data, and upload the source for Lightdash to build on your instance. Around that, content as code expanded to cover dashboards, permissions, AI agents, automations and org roles, and verified content was unified with AI agents so the MCP serves one trusted source.
◆Where it's heading
Two threads are converging. One makes the semantic layer legible to agents - verified content and AI-verified answers share a single source of truth that the Lightdash MCP and outside assistants read from. The other makes the platform something agents can write to, with apps scaffolded locally, built by whatever coding agent the developer prefers, then shipped into a governed instance. The governance framing is carrying real weight in both, since the pitch is that data and metrics stay controlled while authoring moves outside the product.
◆Prediction
Expect the local app workflow and content as code to fuse, so agent-driven changes to dashboards, permissions and apps arrive as pull requests against a Lightdash instance. The pieces are shipped; what these entries do not settle is how agent-authored apps get reviewed or approved before viewers see them.

◆Recent moves

  1. 7d ago

    🤖 Build data apps locally with your favorite agent

    ⚡ SPARK

    Data apps can now be built entirely on a developer's own machine - scaffolded with a CLI, iterated on in any IDE with any coding agent against live data, then uploaded for Lightdash to build on the instance. It moves authoring out of the Lightdash UI while keeping the metrics layer governed, which is the balance the whole Data Apps push has been working toward.

    View source ↗
  2. 11d ago

    📦 More content as code

    Content as code now covers charts, dashboards, spaces, permissions, virtual views, AI agents and data apps, plus scheduled deliveries, alerts, sheet syncs, and even users, groups and custom roles. That puts the whole instance, access control included, into something reviewable in a pull request or refactorable by an agent.

    View source ↗
  3. 11d ago

    SQL Runner: Big Number

    SQL Runner gains Big Number charts, pulling a single value out of a query with an optional label, comparison or trend. Gap-filling work that brings SQL Runner closer to parity with the main chart builder.

    View source ↗
  4. 15d ago

    🎯 Ask for one filter, not every filter

    Dashboards can require at least one filter from a group rather than all of them, with a note telling viewers why. Guided setup aimed squarely at the case where an unfiltered dashboard would be too slow to be usable.

    View source ↗
  5. 29d ago

    🌍 Timezones that just work

    Project timezones now apply across filters, date grouping and displayed timestamps for all Cloud organizations, with per-viewer timezones and the option to pin a chart to a fixed zone. Correctness work that removes a standing source of misread numbers, while untouched instances stay on UTC.

    View source ↗
  6. 1mo ago

    🔌 Data apps can now talk to APIs

    ⚡ SPARK

    Data apps can call third-party HTTP APIs through a Lightdash proxy that injects credentials server-side, with admins restricting base URLs, methods and paths. It turns data apps from things that read the warehouse into things that can act on other systems, without a secret ever reaching the browser.

    View source ↗