Where Colrows is going.
A public view of what we've shipped, what we're building, and what's coming after that. Dates are directional, not commitments.
Generally available - production-grade.
- AI Data Analyst Agent: plain-language → governed SQL with full reasoning chain
- 16+ data sources: Snowflake, Databricks, BigQuery, Redshift, Postgres, MySQL, more
- Compile-time governance: RBAC + ABAC + row/column-level predicates
- Semantic Studio: visual editor for the versioned semantic graph
- Autonomous maintenance: drift detection, conflict resolution, schema-change handling
- Point-in-time reproducible audit log for every query
Federated governance & hybrid deployments.
- Federated semantic governance across multiple business units & subsidiaries
- Hybrid cloud support: shared, dedicated, or fully private deployments on AWS, Azure, GCP
- SCIM provisioning & advanced SSO (Okta, Azure AD, Google Workspace)
- Compliance packs: SOC 2 Type II, HIPAA-aligned deployments, regional data residency
- Public benchmarks for compile-time policy evaluation at 100K+ entity scale
AI Data Engineer Agent & marketplace.
- AI Data Engineer Agent: automated data ingestion, ETL, and orchestration
- Expanded marketplace integrations: dbt, Atlan, Confluence, Notion, Slack apps
- Custom dialect adapter SDK for in-house engines
- Real-time semantic streaming for event-driven AI agents
- Native MCP server for AI agents and copilots
The default semantic layer for GenAI.
- Colrows as the default semantic execution runtime across enterprise AI deployments
- Open semantic interchange - portable graph format across vendors
- Cross-enterprise federated reasoning under data clean-room policy
- Industry-specific reference graphs (BFSI, healthcare, retail, manufacturing)
How this roadmap gets decided.
Three rules shape what ships and in what order. Anything that ships moves from this page to the release notes.
Determinism before features
Every capability rides the same compile-then-execute pipeline. If a behaviour can't be proven at compile time - join path proof, policy predicates, reproducibility - it doesn't ship, however demo-friendly it looks.
Pulled by regulated deployments
Priorities come from production use in pharma, BFSI, and travel retail - not from feature matrices. The most-requested items from customers and wishlist notes get built first.
Autonomy over configuration
The graph maintains itself: autonomous maintenance, drift detection, and conflict resolution. Roadmap items must reduce the human upkeep of the semantic layer, never quietly add to it.
Dates are directional, not commitments. For what's already live, see the release notes; for how the shipped pipeline works today, read how it works.
Tell us what you need.
If something on your wishlist isn't on the roadmap, send us a note. The most-requested items get prioritised.