Asset Bundles
Validate and deploy declarative Databricks bundles per target
Built with Databricks, for Databricks teams
A safe runway from commit to Databricks production.
Fabric Runway is a governed delivery control plane for Databricks. It stages immutable Unity Catalog Volume artifacts, deploys Databricks Asset Bundles and Apps through Fabric Harness, gates every release on Fabric Experiments evidence, and routes each mutation — from a human or an agent — through the same Fabric Platform governance and audit log.
fr deploy
Commit to preview to prod
sha256 digest
Immutable UC Volume artifacts
Evidence gates
Bound to the exact run
Audited events
Humans and agents alike
Fabric ecosystem
Shared foundations across the Fabric family
Augments Databricks, never replaces it
Built around the workloads data teams actually ship
Runway extends native Databricks — it does not duplicate it. Unity Catalog, Asset Bundles, Model Serving, and your workspace stay the authoritative system of record for compute, data, lineage, and governance. Runway drives those native services through the Databricks CLI and adds only what a workspace does not: immutable artifact provenance, release gates, and an audited delivery trail.
Databricks Apps
Ship long-running Apps with scoped compute and identity
Unity Catalog
Content-addressed Volume artifacts and secret scopes
GenAI agents
RAG and agent templates on Model Serving + AI Search
Governed delivery
The governance is the product, not the deploy script
Handlers write nothing directly. A request flows through schema validation, policy checks, and a state machine before any Databricks call — and the event log is appended only after the workspace confirms. Illegal moves have no edge in the machine, so they fail structurally instead of by convention.
See the mutation lifecycleEvery mutation is governed
Deploy, gate, approve, promote, rollback, and secrets sync are Fabric Platform actions with policy, a state machine, and an audit event.
Evidence, not vibes
Promotion requires a fresh Fabric Experiments verdict bound to the exact requested run. Manual and mock verdicts never unlock prod.
Agents are first-class
A Harness agent requesting a deploy goes through the identical gates as a human. The MCP server exposes the same governed actions.
One product, three moves
Scaffold, deploy, and promote without losing the evidence trail
01
Scaffold
Start from a governed template
fr init writes a RAG agent repo wired to a Databricks Asset Bundle, an App, a quality gate, and a pre-built GitHub Actions pipeline.
Explore scaffold02
Deploy
Stage an immutable artifact
fr deploy builds the bundle, uploads it to a content-addressed Unity Catalog Volume path, and requests the release gate — every preview PR gets its own environment.
Explore deploy03
Promote
Turn evidence into production
A reviewer approves; provenance, freshness, and separation-of-duty policies hold; then the trusted worker executes the prod deploy. Rollback redeploys a prior artifact.
Explore promoteThe moat
Six guarantees enforced on every release
These are not lint rules or a checklist — they are policies the pipeline evaluates before a Databricks call is ever made. If one fails, the release stops with a reason you can read, and the attempt is recorded.
Read the evidence chain01
Immutable artifacts
sha256 tree digest, verified after materialization.
02
Provenance gate
Verdicts must come from Fabric Experiments with evidence.
03
Freshness gate
Evidence must be inside the configured TTL window.
04
Separation of duty
The requester can never approve their own promotion.
05
Redacted secrets
Values never touch action parameters, events, or projections.
06
Org-scoped audit
Every row is tenant-stamped; cross-org access is blocked.
The Fabric family
Harness builds them. Experiments judges them. Runway ships them.
Runway does not reimplement a mutation engine, a Databricks deploy SDK, or an evaluation platform. It owns the delivery domain — the catalog, the state machine, the projections, the CLI, MCP, and the console — and delegates the rest to the products that own them.
Fabric Platform
Governs every mutation: durable invocations, policy evaluation, the state machine, and the immutable event log.
Fabric Harness
Owns Databricks transport: UC Volume artifacts, tree digests, bundle validate and deploy, and App activation.
Fabric Experiments
Owns release evidence: versioned datasets, evaluation runs, and the verdict that gates a promotion.
Start in your workspace
Scaffold a governed agent and deploy it today
Install the CLI, run fr init, point it at your Runway control plane, and take one RAG agent from a commit to a gated Databricks promotion.