Blink Agent Builder - You can now vibe code agentic AI apps
by•
Blink is the first vibe coding platform that builds AI agents. Describe what you want — Blink creates an agent that thinks, uses tools, and completes tasks end-to-end. Built-in web search, code execution, vector database, sandbox, and 180+ AI models. We used it to recreate Cursor, Perplexity, and Shortcut in minutes. Now it's your turn.
Replies
good luck with the launch, this is cool!
Blink.new
@bbrooksvc Thank you Brandon! Appreciate your support today!
Can I specify the agent framework as well while using blink?
Blink.new
@pulkit_garg7 Blink apps use the Blink SDK (our own agentic framework). It allows any LLM model, prebuilt / custom webhook tools, context management, stopping strategy, etc. to help you build agentic ai apps.
This looks really good - very curious how the integration layer works? If I build an agent that connects to notion, who owns the Notion oauth application? Does that change based on whether its a persoanl/internal application vs customer facing?
Blink.new
@ryan_brown18 Great question, Ryan. You can ask Blink (coding agent) to build you Notion connection to your own OAuth app if you want; we also are beta testing our pre-made connector with Notion etc. so you can also use Blink's platform Notion OAuth app if preferred. The system is very flexible so really depends on your goals and preference.
I know what I'm testing this weekend - congrats on the product!
Blink.new
@andreitudor14 Thanks Andrei - would love to get your take on the product! If you need any help just message me :) We're always online to help our users succeed!
hmm whats the main difference between Blink and Manus for example?
Blink.new
@colinhz Blink can create a Manus agent for you, if you want. No one else can.
Congrats on the launch, Kai 🚀
The shift from stitching together infra to describing an agent and having it actually work is compelling—especially with built-in context handling and tool orchestration.
From your testing, which part turned out to be the hardest to generalize across different agent types: long-running context, tool reliability, or multi-agent coordination?
Blink.new
@ashokgehlot011 Thank you Ashok! Going from prompt to working agentic software is a breakthrough we've worked hard to achieve. Really glad to see the idea resonating with people!
The hardest part to generalize across different agents - well I think context engineering is always challenging in agentic systems, and that's why we've invested a lot of efforts to provide built-in context management for our users developing agentic systems, so we can make the challenge a solved problem.
Multiagent coordination is actually quite easy, as we are now able to define different agents and call them in the application. Subagents is hard - we are still working on that.
Pandada AI
I’ve played with a lot of AI coding tools, and most of them still feel like fancy chatbots.
This one is different. You’re not writing code or prompts — you’re explaining intent. The agent system does the rest. I rebuilt a Cursor-level app surprisingly fast.
Blink.new
@jeremyrkk Exactly, Jeremy! The experience of using natural language prompts to build cursor-level agentic apps is amazing. Thanks for trying it out and would love to see what you build next with Blink!
Minara
Congrats on the launch. I tried it, and the UI design is really nice!
This looks seriously impressive. The self-correcting + full-stack setup (auth, DB, AI, hosting) in one flow is exactly what builders need right now. Curious to see how it handles real-world edge cases 👀