TLDR: Stream, a leading provider of scalable chat, video, and feeds APIs, has launched Vision Agents, the first open-source, video-first SDK designed for building real-time AI agents. This new platform enables developers to create interactive, multimodal applications that can see, hear, and understand in real time, offering flexible integrations with top AI models like OpenAI Realtime and Google Gemini.
BOULDER, Colo. – October 17, 2025 – Stream, a prominent provider of scalable chat, video, and feeds APIs, today announced the launch of Vision Agents, an innovative open-source, open-platform SDK. This groundbreaking toolkit is set to revolutionize the development of real-time video and audio intelligence within developer applications.
Unlike many existing frameworks that have historically prioritized voice and subsequently integrated video capabilities, Vision Agents were conceptualized and built with a video-first approach. Thierry Schellenbach, CEO and Co-Founder of Stream, emphasized this distinction, stating, “Most frameworks started with voice and later added video. We built the opposite: a video-first foundation that’s open, extensible, and developer-friendly.” This fundamental design choice allows for low-latency video intelligence, crucial for real-time perception and understanding.
The new SDK empowers developers to construct AI-powered agents capable of real-time visual and auditory comprehension, coupled with memory and contextual awareness. This paves the way for a new generation of interactive, multimodal applications across various industries.
Open Platform for Broad AI Integration
Vision Agents are designed for versatility, working seamlessly with Stream Video by default while also supporting integration with other video SDKs. The platform is compatible with a range of leading AI providers, including OpenAI Realtime, Google Gemini, and various custom models. This flexibility ensures that companies can adopt Vision Agents without disrupting their existing infrastructure. Furthermore, current users of Stream Video and Chat will benefit from deep integrations that enhance memory, messaging, and overall performance within their applications.
Core Features for Real-Time Intelligence
The platform’s ability to process live video with minimal latency is a cornerstone of its offering, facilitating real-time perception, sophisticated scene detection, and natural audio or text-based responses. Key features of Vision Agents include:
Video-first intelligence: Optimized for comprehensive scene understanding.
Real-time audio processing: Incorporates transcription, speech generation, and voice activity detection for natural interactions.
Memory and context: Agents can recall details and maintain context across interactions, enabling more natural conversations.
Action-ready design: Built to easily connect with external APIs and services, allowing agents to perform tasks and interact with the broader digital ecosystem.
Diverse Applications Across Industries
The potential applications for Vision Agents are extensive and varied, spanning multiple sectors. These include, but are not limited to:
Manufacturing: For advanced defect detection and quality control.
Collaboration: Enabling AI-powered note-taking and transcription services during meetings.
Gaming: Providing intelligent coaching and dynamic avatars.
Accessibility: Generating real-time captions and descriptions for enhanced inclusivity.
Customer Support: Developing multimodal assistants that can understand and respond to complex customer queries.
Commitment to Open Source and Community Contribution
Stream’s Vision Agents are fully open-source, encouraging a collaborative development environment. The company invites developers and partners to contribute to the platform by extending providers, creating new tools, processors, adapters, and integrations directly on GitHub. Thierry Schellenbach likened the current state of Vision AI to the early days of large language models, stating, “Vision AI today feels like ChatGPT in 2022, it’s just beginning to show what’s possible.”
Also Read:
- Codev Unveils AI Agent Platform to Revolutionize Enterprise Code Generation and Documentation
- Coder Unveils Blink: Accelerating AI Agent Development and Deployment for Developers
For media inquiries, please contact Emily Roberts at Stream.
