September 2026 Summaries

Lovable and Salesforce have partnered to let teams build custom applications and AI agents around Salesforce data using Headless 360, while preserving Salesforce identity controls, permissions, compliance, and its role as the system of record. Lovable apps can read and write core Salesforce objects through individual user accounts, while read-only dashboards and pipeline views can be published directly inside Salesforce with live data; write-enabled apps operate through Lovable’s connector. The integration also brings Lovable into Slack, where users can request apps in natural language, collaborate on builds through Slack Code channels, and create workspace agents for tasks such as quote approvals or account updates. The companies present the approach as a way to eliminate duplicate data, manual re-entry, and separate logins while retaining enterprise controls including SSO, SCIM, security scanning, audit logs, approved connectors, and versioned code. Available immediately, the offering aims to enable business teams to quickly create Salesforce-connected tools and agents that fit their existing workflows.

Lovable has introduced OJ, or Orange Juice, a Rust-based, Vite-compatible development server designed to make its large-scale live app previews faster and less memory-intensive. Unlike traditional Vite instances, which rely on Node.js runtimes and project-installed toolchains, OJ runs as a single binary, reads existing Vite configurations, supports plugins through a compatibility bridge, and uses a Node sidecar only when JavaScript functionality is required. Its architecture is tailored to agent-driven editing by combining file watching, compilation, module graph updates, and hot reloads into a synchronized process that can batch rapid changes rather than showing intermediate states. Benchmarks on synthetic and real-world projects such as Excalidraw and Twenty indicate substantially faster cold starts and memory use ranging from roughly one-third to one-eighth of Vite’s, although comparisons include caveats related to unsupported Vite plugins. Production tests at Lovable showed median preview load time declining from 17.4 to 8.0 seconds and sandbox acquisition falling from 14.5 to 3.0 seconds. OJ is being rolled out gradually to preserve compatibility, requires no changes from builders, and has been open-sourced under Lovable’s GitHub organization for public use and contributions.

Lovable has acquired Sutro, a company whose team spent five years developing a language, compiler, and backend platform designed to make application behavior, rules, security, and reliability more explicit. The acquisition reflects Lovable’s aim to help more people build and deploy trustworthy software without needing to manage all of its underlying complexity. Sutro founder Tomas Halgas and engineers Max Gfeller, Tony Zhan, and Hirad Arshadi will join Lovable, with Halgas leading technical evangelism focused on making the platform’s technology more understandable for builders.

Lovable has launched its Partner Program to connect companies seeking AI-enabled software solutions with independent builders, agencies, consultancies, and systems integrators using its platform. The program includes Expert and Solution Partner tracks, certifications based on shipped work, a directory for prospective clients, referral commissions, and varying levels of support and co-selling opportunities. Lovable says more than 800,000 individuals and agencies already use the platform for client work, while over 500 solution partners, including Deutsche Telekom and KPMG, signed agreements or began certification before launch. Survey findings cited by the company suggest that Lovable has enabled new income and career paths globally, including for people without prior software-development experience, while helping builders undertake a wider range of projects such as internal tools, SaaS products, automations, analytics systems, and client portals. Case studies describe agencies and consultancies using the platform to grow operations, accelerate delivery, and create custom software for organizations ranging from local businesses to government and enterprise clients.

Drafts let users create separate, parallel versions of a project to test changes to layout, design, copy, media, and interactive front-end elements without affecting the live application until the draft is accepted and published. Each draft has its own chat and preview, allowing collaborators such as teammates, clients, or copywriters to explore ideas directly in the project while the owner retains control over what goes live. Users can create unlimited drafts, compare alternatives, accept a preferred version, revert changes, or delete and recreate drafts as needed. The initial release supports projects with or without databases and focuses on front-end changes, while database schema and login configuration changes must still be made in the main project chat; drafts also currently use the published application’s database.