Dyte
Winter 2021AcquiredReal-time audio & video SDKs, ready to launch 🚀
Pressure-test this opportunity
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Real-time audio & video SDKs, ready to launch 🚀
Explore the risks and possibilities with a prompt for ChatGPT, Claude, or your agent.
Dyte sold the layer between a developer's product and the hard parts of a live call. Its SDKs handled participant state, audio and video, roles, layouts, chat, polls, recording, live streaming, and interactive plugins. A customer could ship a branded meeting experience without building a media stack or accepting the fixed interface of a consumer conferencing app.
Abhishek Kankani and Palash Golecha started the Bengaluru company in 2020 and entered Y Combinator's Winter 2021 batch. Dyte raised an $11.6 million seed round, grew beyond fifty employees, and said its infrastructure had reached hundreds of thousands of people. In June 2025 the whole team joined Cloudflare. The product became Cloudflare RealtimeKit, and the old Dyte SDKs entered maintenance mode.
The transaction price is not public. This was nevertheless a clear strategic sale: Cloudflare supplied a global SFU, STUN, and TURN network, while Dyte supplied the client SDKs, UI components, and developer journey that made that infrastructure usable.[1]
Image 1 / 1
Kankani and Golecha began building in November 2020, when remote work and online classes had made video a default product surface. Their first idea was a standalone calling application. The turning point came when customers asked to embed the technology inside their own products. Dyte moved down the stack and sold an API and SDK instead.[2]
The first implementation carried a revealing debt. A client wanted the founders' demo application as an SDK, so the tiny team adapted it. UI elements and business logic remained tangled. That helped Dyte ship, but made customization and maintenance harder. The company later rebuilt the product around Web Core, a lower-level state and media SDK, with UI components layered on top.[3]
That progression defined the company. Dyte was not only renting media servers. It was converting repeated integration work into primitives that a customer's engineering team could control. Kankani described the product as something its developer-founders would themselves want to use: fast to integrate, well documented, and open to customization.[4]
Dyte's basic unit was a meeting that a developer created through an API and joined through a client SDK. Presets defined participant roles and permissions. UI Kit supplied setup screens, participant grids, controls, chat, and polls; Core SDK exposed state and media for teams that wanted their own interface. Recordings, RTMP broadcasts, webhooks, analytics, and plugins extended the call.
The product spanned web, React, mobile, and desktop environments. TechCrunch showed Dyte changing brand colors, visible controls, roles, and plugins in minutes. Plugins could add a shared browser, whiteboard, file viewer, or game without reducing the meeting to screen sharing.[7]
Operations remained difficult behind the polished SDK. Dyte's first recording system launched after a two-week build. A Chrome bot joined each meeting and buffered media in memory. It failed under production load on the first day, exhausting memory and crashing jobs. The team replaced it with a redesigned pipeline.[8] The story matters because it shows what customers were buying: not an API wrapper, but accumulated fixes for browsers, devices, networks, recording, and state.
Dyte charged by participant minute. In 2022 the first 10,000 minutes per month were free; recording and RTMP broadcasting cost extra. This usage model lowered the cost of a pilot and aligned revenue with calls, but also tied gross margin to media consumption and infrastructure efficiency.
Dyte targeted software companies that needed real-time audio or video inside their own workflow: telehealth, education, hiring, fitness, social, gaming, and events. The buyer was an engineering or product team for whom a generic meeting link broke context, branding, permissions, or data flow.
The company sat between raw WebRTC and a finished conferencing product. Raw WebRTC offered control but left signaling, NAT traversal, media routing, reconnection, observability, and cross-platform behavior to the developer. Zoom and Meet offered reliable calls but controlled the experience. Dyte promised a working middle layer.
The category was crowded. Twilio, Agora, Vonage, Daily, 100ms, LiveKit, and others sold combinations of global media infrastructure and client tooling. Open-source WebRTC components reduced the cost of starting, while large networks could spread SFU and TURN investment across many products.
Dyte differentiated through developer experience, fast integration, component-level UI control, and interactive plugins. Those features were legible to a product engineer. They were harder to defend as an independent infrastructure company once better-capitalized platforms improved their client layers.
Dyte used usage-based infrastructure pricing with a free allowance. It could expand within a customer as call volume, recording, and broadcasting grew. Enterprise requirements such as data residency, support, and negotiated scale pricing added another path to larger contracts.
The model also carried two costs. First, media delivery has real variable expense, so price competition can compress margin. Second, every supported browser, operating system, device, framework, and network condition creates maintenance work even when usage is low. Cloudflare could absorb both burdens into a network serving many other products.
The strongest public numbers came from the company. Kankani reported twenty-fold growth during 2021 and 300,000 people across more than 300 cities using Dyte-powered products by year end.[9] By April 2022 he put the figure at 500,000 people and said the team had grown from three to more than thirty in eighteen months.[10]
Dyte's own timeline later recorded a fifty-plus-person offsite and more than 1,000 upvotes for a Product Hunt launch.[11] These signals show adoption and organizational growth, but not recurring revenue, retention, gross margin, or the number of paying accounts. The acquisition itself and Cloudflare's decision to retain the full team are stronger evidence of technical value than a disclosed independent business outcome.
Cloudflare had launched Calls as programmable WebRTC infrastructure. Kankani sent a cold message, the companies began working together, and the relationship lasted about a year before Cloudflare proposed joining forces. Kankani said Dyte had received acquisition interest before, but this combination came with trust already established.[12]
Cloudflare described the fit directly. Its global network handled media routing. Dyte brought SDKs, UI components, and developer experience. Buying the team closed the gap between low-level infrastructure and a product that application developers could integrate.
Dyte had already expanded from four media regions to eight. Cloudflare operated real-time infrastructure across hundreds of cities. The buyer could improve latency and unit economics without forcing Dyte to fund a comparable network. In exchange, Dyte ceased to be an independent vendor.
This was not a shutdown caused by an unwanted product. The whole team joined Cloudflare, Kankani became Head of Emerging Technology and Incubation India, and customers continued to receive support. Acquisition terms were not disclosed, so the financial return cannot be judged.
The name changed, packages changed, and old SDKs entered maintenance mode. Cloudflare's migration guide tells Dyte customers to move to RealtimeKit.[13] RealtimeKit now combines UI and Core SDKs with Cloudflare's media network, recordings, transcription, plugins, and webhooks.[14]
The product survived, but customers still own the release risk created by browser changes, mobile SDK versions, networks, and vendor migrations. Cloudflare's current release notes include breaking APIs, Safari failures, initialization retries, and media-connection error handling.[15] That recurring uncertainty is the opening for a new company.