Agenta

If you're looking for an open-source alternative to dots/grokbot, without vendor any lock-in, and fully customizable. checkout @agenta_ai github.com/agenta-ai/agenta

this simple trick makes a big difference on how useful your agents: stop thinking of skills as something fixed, each time you use them give the agent feedback and ask it to improve it

Here is an awesome trick to help reviewing PRs Add this prompt to your agent "Download the skill from in github.com/cathrynlavery/dia… and use it to explain the architecture changes in the PR"

Here's the new @agenta_ai. It's an open-source alternative to Claude Cowork -- but you choose your own harness (claude, codex, pi) and model (any, including self-hosted). You can use it to build AI coworkers and automations. You build agents by talking to them. They learn from your feedback and edit their config. demo:

Some things take days. Some take years. Last year our team sat down and sketched an idea that felt pretty far away. Today it’s not. We’ve been building quietly at @agenta_ai, and next week we finally get to show it. Launch Week #2 starts Monday.

925 days ago, I wrote the first line of @agenta_ai Now we are used by hundreds of teams. But our brand still looked amateurish. Today we fix that. Hope you like it.

Starting a video series on LLMOps First topic: LLM Observability and why do you need it?

How do you know if your RAG system actually works? Most teams spend months optimizing but can't tell if their changes help or hurt performance. I collaborated with Karkar Nizar on a guide covering the key metrics Full blog post in the last tweet in the thread 👇

🔒 Launch Week Day 5: Agenta is now SOC2 Type 2 compliant! Not the flashiest announcement to end our week, but crucial for teams building AI in regulated industries or with sensitive data. 🧵 A quick thread on our Launch Week...

Today for Day 3 of Launch Week in @agenta_ai : The Prompt & Configuration Registry A simple place to manage all your LLM prompts with proper version control. Teams kept losing track of which prompts worked best, so we built this.

📣 Launch Week @agenta_ai Day 2: Custom Workflows Connect your LLM app to Agenta and let your team iterate on the full configuration without touching code.

Launch Week Day 1 Excited to announce Model Hub – now you can connect ANY LLM to Agenta! Azure, AWS Bedrock, self-hosted models, fine-tuned models... if it has an API, you can use it in Agenta. 🚀

Next week is Launch Week at @agenta_ai! While I’ve been sharing product announcements, I’ve never introduced the incredible faces behind our work. We’re a team of 7 spread across three continents, mostly working remotely. This is genuinely the team I’m most proud to have ever worked with. We spend more time with each other than anyone else, and building something meaningful together has been incredibly rewarding. Though we haven’t all met in one place yet (we will soon!), last week some of us gathered for a pre-launch meetup, and it was fantastic.

We're doing our first ever Launch Week starting Monday! 🚀 For 5 days straight, we'll be sharing new features we've been quietly building.

New Tutorial: Building test sets from production traces for LLM eval in @agenta_ai When your LLM app is in production, your users find edge cases you never thought of. Here I show how to: - Convert (batch) traces → test sets - Create test sets in the playground