What happened
Lovable announced four new connectors: Perplexity, ElevenLabs, Firecrawl, and Miro. The company presents them as ways to build applications that can research, speak, pull live information from the web, and turn a visual planning board into a working prototype. The source is an announcement published on December 16, 2025.
The Perplexity connector is positioned as a way to put an AI research assistant inside an application. Lovable says an app can ask Perplexity a question, receive a synthesized answer, and show sources behind that answer. Its examples include a plant-care app explaining a problem with tomatoes, a recipe tool suggesting substitutions, a travel planner checking current conditions, and a sales tool preparing a team before calls. The important product boundary is not merely a search box: the application can request research for its own user-facing workflow.
The ElevenLabs connector adds speech and sound-generation capabilities. Lovable describes examples including a children's app reading stories in different character voices, a language-learning tool handling pronunciation, a meditation app pairing voice with a soundtrack, a game producing background music, and a daily briefing read aloud. The announcement says ElevenLabs supports natural-sounding speech in dozens of languages and voices designed by the user; it also refers to ElevenLabs music generation.
Firecrawl supplies web extraction. Lovable describes Firecrawl as turning a webpage into clean, structured data, including product information, job listings, pricing tables, and documentation. Suggested uses include aggregating job listings from Greenhouse, Lever, and Workday; tracking competitors' price changes; collecting venue calendars; and scraping API documentation for a lead-scoring tool. Lovable says Firecrawl was free for Lovable users through January 2026 when they opted for a managed instance.
Finally, the Miro connector lets a Lovable agent use the content of a board: flows, clusters, sticky notes, structure, and other planning material. Lovable frames the resulting workflow as moving from a morning ideation session to an afternoon prototype, or providing a board link so that the agent can use the board's planning context while building a landing page or app. The stated setup path for all four is Workspace Settings, then Connectors, followed by connecting the selected service and describing the desired feature.
What the archive adds
This is a new fact on a known connector theme. The archive records an earlier Telegram post about Lovable making Parallel available for web search in generated applications; this announcement broadens that capability set to four providers and adds voice, extraction, and Miro-board inputs. There is no genuinely related prior Field Note to link.
Trend signal
This supports Agent-ready software exposes capabilities. Lovable is packaging distinct external functions—research, voice, web extraction, and visual planning context—as selectable capabilities an application-building workflow can compose. The announcement does not establish how permissions, data handling, or operational controls work, so it supports the direction more than the safety claim in the pattern.
Connections to other events
The closest related archive event is Lovable's earlier Parallel integration for web search in generated apps. The new announcement makes the product story broader: search is one connector category alongside extraction, media generation, and a planning-workspace input. Perplexity is also appearing as a capability inside other products, rather than only as a standalone research destination.
Why it matters
For teams using AI app builders, the practical change is a shorter route from a product idea to an application with external capabilities. A prototype can incorporate current web research, extract structured material from pages, generate an audio experience, or start from a team board without first implementing each integration from scratch.
That does not mean the integrations are interchangeable. Perplexity is described as research with sources, whereas Firecrawl is described as extraction of structured page content. One can support an answer-oriented interface; the other can support a pipeline that collects pages for a dashboard or workflow. Miro is different again: it is context for app construction, not a source of live external facts. Teams should choose the connector based on the job their application needs to perform and define how outputs will be checked before users rely on them.
What happens next
- Lovable will expand its connector catalog or the scope of existing connectors. (6 months, 70%). Watch for: a named connector release or expanded connector controls.
- Builders will combine research or extraction connectors with application-specific user flows. (6 months, 60%). Watch for: examples or templates that join connectors in one Lovable application.
Limits and uncertainty
The announcement is a product description, not an independent evaluation. It does not specify connector pricing for Perplexity, ElevenLabs, or Miro, usage limits, supported regions, data-retention terms, authentication mechanics, or the quality and reliability of outputs. Its Firecrawl promotion is explicitly time-bounded through January 2026 for managed instances, but the source does not establish what follows that period. The examples illustrate possible applications; they are not evidence that every use case will work reliably or meet production requirements.
Skill to build now
Build one small workflow that distinguishes research from extraction. For example, make a prototype that uses Perplexity to answer a user question with cited sources, while using Firecrawl only to collect a defined set of fields from a specified page. Write down the acceptance check for each output: source visibility and factual review for the answer, schema completeness and page-level validation for the extraction. This forces the product design to match the connector's stated role.
Reader takeaway
Lovable's update is a capability expansion, not a general claim that an app builder can safely automate every external task. The useful change is concrete: research, voice, web extraction, and Miro context can now be added through its connector workflow. Treat each as a different product primitive with its own verification requirement.