Unexpected talk 😅 /cc Upsonic Mumtaz Vural bertug

We held a great event where the founders of AI agent startups came together. It was the first event we made at Upsonic . Our aim was that everyone could meet each other, so we accepted guests only with an invitation and kept the group small. This created a warm and friendly

Onur Ulusoy just sat down with the miakassissa team for a deep-dive YouTube interview on agentic AI, Upsonic’s reliability layer, and how enterprises can navigate a full agentic transformation. 🎙️ What you’ll learn in 40 minutes: 1- Why “agentic AI” matters now – and

This is a commitment photo. We have come together to become the most trusted AI agent partner in the fintech industry.🦾 Every day, our team focuses on enhancing agent reliability and ensuring AI safety. Each month, new companies launch with our agents in production, with our

🚀 Upsonic v0.66.0 is live! We shipped these in this version: ✨ Expression Language (UEL) - Build AI pipelines with pipes 🔄 StateGraph & Recursive Streaming - Smarter agent workflows 🧪 MCP Integration Tests - Rock-solid protocol support 🐛 Bug fixes and doc updates!

Bugün enterprise’da AI agent geliştirme sürecini konuşuyoruz: PoC’den canlıya giden gerçek agentic dönüşüm sürecini adım adım anlatacağız daha sonra agentların safetysini sağlamak için tasarladığımız güvenlik politikalarından bahsedeceğiz. Şirketlerde yapay zeka projeleri

all we need is… (🧢 Upsonic)

Enes Karatas just shared a UCP Shopping Agent example. With UCP, an agent can build a complete setup end-to-end, it doesn’t just pick one item. It lists everything you’d expect in a real setup: a keyboard, mouse, mousepad, and the rest of the essentials. We’ve made it easy to

v0.70.0 is live 🚀 This release is all about cleaner agent runs, better session structure, and more control when things get long-running. What’s new in v0.70.0 - AgentRunInput & AgentRunOutput: Dedicated input/output classes for agent runs so your code stays organized and

🚀 New Feature: Culture Agent behavior just leveled up. No character breaks. No spaghetti prompts. Just real behavioral boundaries. • Agents internalize their role • Out-of-scope requests get handled gracefully

We cooked something for agents that do everything without asking permission. ex: you have a customer support agent, but if a user asks it to write some code, your agent just says yes. The Culture feature in our Python SDK prevents that. docs.upsonic.ai/concepts/cultu…

🛡️ Safety Engine Series #1: PII Anonymize Policy Stop sensitive data leaks and action errors. No exposed customer data. No compliance violations. Protected agents. • Use prebuilt PII policies or build your own • Prevent LLMs taking wrong actions on sensitive information

🔗Team Modes Series #1: Sequential Mode Turn your git log into a ready-to-post tweet. Two agents. Automatic context handover. No glue code. Agent A reads commits, filters noise, extracts what matters Agent B writes a dev-native tweet — no hashtags, no corporate tone 👇🏻

DevOps Agent on Telegram 🤖📱 Your server, in your pocket. No SSH. No dashboards. Just text it. OpenClaw-style autonomy: observe → decide → act 🔍 Reads logs, tells you what's wrong 🧪 Runs commands in a sandboxed workspace 📝 Behavior defined in markdown, not code

x.com/i/article/2039…