LindyAI

LindyAI, also known as Lindy, builds AI assistants and a no-code platform for creating agents that automate work. Founded by Flo Crivello, its products span email and calendar assistance, business workflows, agents that operate cloud computers, web-app creation, and phone calls. Its February 2026 Lindy Assistant announcement brought inbox triage, scheduling, meeting preparation, notes, and follow-ups into iMessage, alongside email, Slack, and a browser.[8][3][4][5][7]

Lindy’s organizational and technical origins preceded its assistant products. Crivello previously founded Teamflow, a virtual-office service where video avatars moved around a two-dimensional space and participants could hear nearby colleagues. On July 26, 2021, Teamflow announced that it had closed a $35 million Series B, led by Coatue Management with Menlo Ventures and Battery Ventures participating, bringing reported cumulative funding to $50 million.[9]

In an interview published in July 2025, Crivello recalled using GPT-3 to summarize Teamflow meeting recordings. The sales team then asked for automatic Salesforce updates. The work expanded from extracting information for Salesforce fields to supporting HubSpot, other CRMs, and tasks in Linear. Crivello also recalled early contact with LangChain’s Harrison Chase as the agent idea developed.[8]

Crivello said Teamflow’s growth flattened as people returned to offices, and that he decided to pivot to Lindy in early 2023. The transition involved letting go of two-thirds of a team of more than 50, including the entire go-to-market team. He described Lindy as a pivot within the same legal entity. A Twitter demo combining working and Figma prototypes attracted a waitlist of 70,000 people, he recalled. The March 24, 2023 Latent Space episode introduced him as the founder of the newly launched AI assistant startup.[8][1]

The March 2023 assistant announcement presented calendar scheduling, email triage and reply drafting, meeting recording, and recurring workflows triggered by emails or calendar invitations. One concrete workflow gathered previous email exchanges and meeting recordings with upcoming attendees, then sent a briefing in Slack five minutes before the meeting. Lindy could join Zoom, Google Meet, or Teams meetings and use their content for follow-on tasks such as updating Salesforce.[2]

Lindy’s early technical approach included writing code to execute requests such as booking a meeting or sending an email. The company described guardrails that considered both the model’s perplexity before an action and the potential impact of getting it wrong when deciding whether to proceed or request confirmation.[2]

Actual behavior fell short of the “AI employee” promise, according to Crivello’s later account. In one recruitment workflow, users asked Lindy to find 50 software engineers and email them about joining a startup. Instead of a recruitment message, recipients received text such as “the user wants me to send an email to 50 software engineers.” Crivello recalled poor retention and surprisingly forgiving early users.[8]

A prospective customer’s detailed requirements helped prompt a rebuild: Crivello concluded that the existing architecture could not reliably deliver what customers needed and that waiting for better models would not suffice. Lindy 2.0 adopted a familiar boxes-and-arrows interface for its agent builder and positioning as the “Zapier of AI.” Crivello credited YouTuber MattVidPro with trying the quietly released version and publishing a video that accelerated growth.[8]

Lindy 3.0, presented in a July 2025 announcement, expanded agent creation, capability, and team deployment. Agent Builder created custom agents from prompts. Autopilot gave agents cloud computers they could see and operate, extending their reach beyond API integrations. Team Accounts provided centralized management, sharing, and monitoring. The company reported more than 6,000 integrations at that time.[3]

The August 2025 Lindy Build announcement extended the product into generating web apps with databases, integrations, and frontend and backend configuration. The company described testing through interaction: Build clicked through the app’s forms, interfaces, and features to find and repair problems, rather than only reviewing code. Apps also connected to Lindy Agents for work such as prospecting, marketing, and support.[4]

In October 2025, Lindy announced Gaia, a phone agent powered by Deepgram Flux. Its uses included inbound customer support and receptionist work, such as screening calls, scheduling appointments, and escalating to team members. Deepgram described real-time actions including Salesforce or HubSpot updates, note logging, and Slack notifications.[5][6]

Flux integrated turn detection into speech recognition. In Deepgram’s partner announcement, Reuben Dubester, identified as ML Engineer & Cognitive Architect, described it as a near drop-in replacement for Lindy’s existing voice architecture. He said its state machine reduced complexity Lindy had been handling for turn-taking and speculative response generation.[6]

The February 2026 Lindy Assistant combined those familiar administrative tasks in a conversational interface: users could text requests such as preparing for a meeting or scheduling a follow-up. Meeting preparation drew on past emails, previous meetings, and calendar context. For email replies, the described workflow retained a human step: Lindy drafted, and the user reviewed and sent.[7]

In the interview published July 17, 2025, Crivello confirmed the host’s $52 million cumulative fundraising figure and explained that much of the money had been raised for Teamflow. This was financing across the continuing legal entity, rather than an announcement of a separate Lindy round.[8]

In that interview, Crivello reported “high seven figures of revenue,” 35 employees, and “five or eight contractors.” He did not specify the revenue period, currency, or recurring status. He described typical users as small businesses and startups of roughly five to 100 people, mostly in technology.[8]