Lovable apps can search the live web with Perplexity
A managed connection needs no Perplexity key and bills search in Lovable credits.
Every Lovable update in one feed — release notes, features and changes, straight from the source.
40 updates
A managed connection needs no Perplexity key and bills search in Lovable credits.
Paste a URL or upload a ZIP, and the first import is free until a working preview, though backends, sign-in, and passwords stay behind.
Set a goal such as your first 100 signups and Lovable suggests a plan, then creates campaigns on Google, YouTube, Maps, Facebook and Instagram.
The example combines Claude Opus for visual identification and reasoning with Claude Sonnet for fast web search and synthesis.
Lovable says the new model scores 6 to 12% higher than GPT-5.6 Sol on its own app-building benchmark, with bigger gains in intent alignment.
Opus 5.5 matches Opus 5's build quality in Lovable while using up to 48% fewer steps, per Lovable's internal benchmarks.
Freelancers and agencies building client work on Lovable can now get certified, join a partner directory, and earn more.
The Series C, led by Menlo Ventures, follows a run of monetisation, security, and governance features since December's Series B.
The Trust Center at /.well-known/trust.html checks HTTPS, dependencies and access controls with no setup and nothing to fill in.
Lovable says the model scores highest yet on shipping complete, working apps, with less run-to-run variance than Opus 4.7.
End users link their own Drive, Calendar and other tools for a version personalized to their data and permissions.
Builders can now pick and switch between four new models under Connectors, AI to power their app's AI features.
Any published Lovable app can now expose its actions to AI assistants, with Lovable auto-generating and hosting the OAuth-protected MCP server.
Live now with nothing to enable, the icon changes when Lovable is building, needs input, or finishes a task.
Lovable says the model does more per request and flags risky prompts better, improving security.
The beta feature scans code and live error logs for issues like failed sign-ins or payments, and is limited to Pro, Business, and Enterprise workspaces.
Generated apps can now include voice input and output, from voice assistants to hands-free tools.
Chat about a Lovable project, export a Claude Design straight into it, or ship Claude Code changes to a live app.
Lovable demonstrated the feature by building and launching its own merch store.
The switch aims to improve discoverability in search engines and AI answer engines like ChatGPT and Perplexity.
The 10-15 second scan checks for database misconfigurations, missing RLS policies, and authorization gaps.
Pulls call notes like feature requests or bug reports straight into a Lovable project prompt.
Subagents handle research, review, and QA in parallel, with each one's work traceable in the activity log.
Lovable says the model works well for agentic features in apps built on the platform.
New connectors let apps built in Lovable read and write Gmail, Calendar, Drive, Sheets, Slides, Maps, and BigQuery data.
Save markdown-based playbooks that Lovable loads automatically, personal or team-wide, with five built in and more writable on request.
Tyler Bruno's ground-up explainer covers how Skills work and how to write one well, for builders on Lovable or any AI tool.
New builds get full SSR and unique social previews, while old ones get automatic pre-rendered HTML snapshots at no extra cost.
The new Aesthetics layer lets users set typography, layout and colour preferences and preview concepts before Lovable generates the app.
Entrants build a web app for a mom in their life using Lovable's AI app builder.