According to Retool's changelog, the company shipped stable updates across its self-hosted platform line on September 9, bundling an upgraded Microsoft SQL integration v3.0 with a centralized AI Visibility portal designed to track model tokens and credit consumption. For teams running Retool on their own infrastructure, this consolidates database plumbing and AI spend observability into a single release cycle — the kind of control-plane work that decides whether a deployment scales or quietly bleeds margin.
Microsoft SQL v3.0 — the real lever
The v3.0 upgrade targets what most enterprise Retool shops actually run behind their custom business applications: a Microsoft SQL backend feeding internal tools, admin panels, and operational dashboards. The changelog frames it as a stable release, which closes out the v2 maintenance window and pushes the migration question onto internal platform owners.
The operational calculus is straightforward. A major MSSQL connector version bump typically brings query handling, connection behavior, and reliability improvements that cut down on the runtime overhead teams absorb when Retool apps hammer a heavily loaded SQL backend. For shops that have standardized internal tools on Retool against MSSQL, this is the version worth benchmarking against current workloads before promotion. For everyone on Postgres or MySQL, this piece is largely irrelevant to the database layer — though the rest of the release still applies.
AI Visibility — tokens and credits under one roof
The second piece is the strategically interesting one: a centralized portal for tracking AI model tokens and credit usage across the enterprise deployment. Until now, most Retool admins have had to reconcile AI spend through third-party provider dashboards, internal tagging, or manual log scraping. A native visibility layer inside the admin console removes that overhead.
For finance and platform teams, this shifts the unit economics of building AI features inside Retool apps. Instead of attributing LLM costs through after-the-fact reconciliation, admins can see token consumption tied to specific apps, environments, or workflows directly in the platform. That is the prerequisite for internal chargeback models, budget enforcement, and the cost discipline that prevents an AI feature prototype from becoming an uncontrolled line item.
The "credit usage" framing suggests Retool is also surfacing spend tied to its own bundled AI features — meaning admins no longer need to reconcile third-party API spend separately. If those numbers reconcile cleanly against actual invoices, the administrative overhead of running AI-augmented internal tools drops measurably.
The bottom line
Self-hosted Retool shops running MSSQL backbones should schedule a v3.0 migration window and benchmark against current workload behavior before promoting it. Every enterprise customer — regardless of database choice — should activate the AI Visibility portal on day one. Token and credit tracking at the platform level is now table stakes for any serious internal AI deployment, and Retool shipping it natively removes one of the more persistent administrative debt items most platform teams have been carrying.
For organizations comparing Retool against alternative visual engineering platforms for custom business software, this release does not change the fundamental calculus — but it does signal where Retool is investing: operational tooling enterprise customers actually need, not just new feature toggles.