SAN FRANCISCO, October 14, 2026 – Builder.io today launched Agent-Native in beta, a free, open-source TypeScript framework for building agentic apps. An agentic app is an AI agent with the intuitive interface of an app. The agent and UI work as one, sharing the same actions, data, and application state. Anything a person can do in the app, the agent can do too.
AI agents make it possible to describe an outcome and delegate the steps, but chat interfaces alone make it difficult to inspect results, refine work, and collaborate. An empty prompt box also leaves people guessing what an agent can do, while powerful AI workflows often remain limited to the individuals who created them. Many workflows need intuitive interfaces that make an agent’s capabilities visible and let people see, steer, and refine its work.
Agent-Native provides developers with an open foundation for building AI agents with interactive UIs. People can delegate work through natural language, see and refine results directly in the interface, and take control whenever they need to, without being tied to a single AI model or proprietary platform.
“Every kind of work has an interface that helps people do it, and agents shouldn’t take that away,” said Steve Sewell, CEO of Builder.io. “With Agent-Native, the agent and the app are equal partners. You can hand work to the agent, see the result right away, and take control whenever you need to.”
One Framework, Shared Actions Across Six Surfaces
Agent-Native connects the agent and UI through three shared foundations: actions, data, and application state. When a person makes a change in the UI, the agent has access to the updated information and relevant context. When the agent completes work, the results appear in the app for people to inspect, refine, and share.
Rather than building separate functionality for agents and user interfaces, developers define each action once. Agent-Native makes those actions available through six surfaces: the UI, agent, HTTP API, CLI, Model Context Protocol (MCP), and Agent-to-Agent (A2A) protocol, using the same validation, permissions, and implementation.
Unlike computer-use agents that navigate software by clicking through interfaces, agents built with Agent-Native call the same actions the UI uses. This approach can reduce unnecessary steps and token consumption while enabling agents to complete work directly. Agents can also run on schedules, respond to events, and work with connected tools.
An Open Foundation for Agentic Apps
The Agent-Native beta includes:
-
Shared actions across six surfaces: Define capabilities once and make them accessible through the UI, agent, HTTP API, CLI, MCP, and A2A.
-
Connections to tools and other agents: Connect external services through MCP and expose capabilities through A2A, with advanced agent-to-agent connections in alpha.
-
Built-in agent and team capabilities: Instructions, skills, memory, scheduled tasks, authentication, permissions, and collaboration features.
-
Support for multiple AI models: Bring an API key for Anthropic, OpenAI, Google Gemini, OpenRouter, or any OpenAI-compatible endpoint, or run a local model with Ollama.
-
Optional Agent-Native Toolkit: Ready-made building blocks for editing, collaboration, version history, and application setup.
Builder.io has also built three open-source agentic apps using the framework: Clips for capturing and working with screen recordings and voice notes, Slides for creating and refining presentations, and Design for generating and editing interactive designs. These apps, currently in alpha, demonstrate how Agent-Native can power different workflows, with agents and intuitive interfaces working together.
As AI makes software easier to inspect, copy, and rebuild, competitive advantage increasingly shifts toward how quickly teams can build and customize tools for their own workflows. By making Agent-Native free and open source, Builder.io gives developers a foundation they can extend, personalize, and deploy using their own models and infrastructure.
Availability
Agent-Native is free, open source, and available in beta today.
Developers can start locally with the TypeScript framework using:
npx –yes @agent-native/core@latest create my-agent –standalone –template chat
Agent-Native supports PostgreSQL databases and deployment to Node.js environments and platforms including Vercel, Netlify, Cloudflare, and AWS. Developers can bring their own AI models, tools, and infrastructure while retaining ownership of the applications they build.
Teams can also build agentic apps online with Builder.io, with building, hosting, database, authentication, and deployment handled by the platform, without local setup.
-
Website: https://www.agent-native.com
-
Documentation: https://www.agent-native.com/docs
-
Source code: https://github.com/BuilderIO/agent-native
-
Launch announcement: https://www.builder.io/blog/introducing-agent-native
About Builder.io
Builder.io is a collaborative platform for building with AI agents. Teams use Builder.io to build, review, and ship software, from agentic apps with rich, intuitive UIs to new projects and features in existing codebases. The company also created Agent-Native, the free, open-source TypeScript framework for building agentic apps. Builder.io is based in San Francisco. Learn more at https://www.builder.io.
Media Contact
Company Name: Builder.io
Contact Person: Geoffrey Taft
Email: Send Email
Phone: 415-275-0615
Address:95 3rd Street, 2nd Floor
City: San Francisco
State: CA
Country: United States
Website: https://www.builder.io/