hey hn, we've been building the web infrastructure for ai agents at tinyfish. our agent primitive was already out. just went live with 3 new primitives - search, fetch, and browser - all under one api.
the team wrote some details on how they built fetch and browser (search deep dive coming soon). we forked chromium to build a browser where automation runs below the js sandbox. clicks go through the native trusted event path, dom reads happen in native code with zero javascript executed. separately, we built a fetch that renders pages in a real browser and returns structured content. because status 200 with 14 characters of text isn't actually a successful fetch.
There's no cap per company, depends on the business viability and outlook. Robin wants to flexible with the amount and type of investment. So it can be $250K, $1M, or even the entire $2M to one company!
Yes, it's another accelerator announcement. But hear me out!
I am from tinyfish.ai (web agents + browser infrastructure for autonomous web workflows like navigation, auth, state, operations on the live web). We partnered with Mango Capital to put $2M behind founders building agentic applications on top of it.
Partnered with 15+ companies to give you the entire stack — MongoDB, v0 by Vercel, ElevenLabs, Fireworks.ai, Google for Startups, Composio, and others — providing free credits and direct channels to their eng teams (not their support teams).
The part that might actually interest you:
The application is a working MVP. You build an app, record a demo video of your code running and app working, and post the video publicly on X. (No slide decks please, everyone’s had enough of them).
We're specifically not interested in:
- Chatbots with a nice UI
- "AI agents" that make one API call and format the output
- Anything that can be built without browser infrastructure
We are interested in apps that do actual labor on the live web — multi-step workflows, real session management, complex UI navigation, and automate ops at scale.
9 weeks, fully virtual, rolling admissions from Feb 17. $2M seed pool, funding decisions made directly with Robin Vasan (Mango Capital).
Do check it out if that sounds interesting: tinyfish.ai/accelerator
We are the engineering team behind Mino. We are building the infrastructure layer to help developers scale web agents from demo to production.
Here's the problem: Building an agent is easy; scaling it is expensive. Relying purely on inference loops (screenshot → reason → click) creates a flat cost curve.
It never gets cheaper or faster, no matter how many times you run the same task.
What we did:
Convergent Infrastructure Mino is a headless runtime designed to optimize agent execution over time.
Instead of executing every run as a blank slate, Mino treats navigation as a convergence problem. As your agent repeats tasks, our infrastructure analyzes the successful execution traces to optimize the path.
Early Runs: Prioritize adaptability and exploration (handling dynamic DOMs/UIs).
Later Runs: Prioritize speed and determinism.
The result is an API that warms up like a JIT compiler, becoming faster, cheaper, and more robust the more traffic you push through it.
The Primitive: We expose this as a stateless, async API designed for composition.
the team wrote some details on how they built fetch and browser (search deep dive coming soon). we forked chromium to build a browser where automation runs below the js sandbox. clicks go through the native trusted event path, dom reads happen in native code with zero javascript executed. separately, we built a fetch that renders pages in a real browser and returns structured content. because status 200 with 14 characters of text isn't actually a successful fetch.
details in the posts. browser: https://www.tinyfish.ai/blog/building-a-browser-for-the-agen...
fetch: https://www.tinyfish.ai/blog/production-grade-web-fetching-f...
please do try them out at tinyfish.ai, 500 credits are on us.