The AI Product Development Platform
TL;DR: Today, we’re launching Vellum for Agents; an agent builder that takes a plain description of what you want to automate and asks follow up questions to understand your needs. It connects to your tools, handles the logic, and shows you exactly what the agent is doing so you are never guessing why or how it works. To celebrate our launch we’ll be giving away free credits for everyone who signs up today (worth $100). Launch video here.
We’re a W23 company with deep experience running AI in production. We’ve built AI tooling used by teams at Drata, Redfin, Seeking Alpha, Headspace, and Swisscom to launch and operate real AI solutions.
That experience led us here.
Today, we’re sharing a new version of Vellum, an agent builder for ops teams. You describe the task you want to automate, and Vellum gives you a working agent in minutes. It asks follow up questions, connects to your tools, handles the logic, and shows you exactly what the agent is doing at every step, so there is no guesswork.
Common problems on using AI agents
In 2025 I watched as our engineering team gained superpowers. Coding agents like Devin, Claude Code, Cursor seemed to move them to a whole new level. By end of year, 40% of the work was done by agents and their productivity is up 2.5x.
That impact never really reached the rest of the company.For those of us who are not engineers, getting started with agents is still hard.
While the models are good enough to help with Ops, Finance, Sales, and Marketing tasks, there are frequent challenges that we’ve observed:
- Too many off-the-shelf tools that almost work, but break in specific edge cases
- Prompts you tweak over and over, but still do not fully trust
- Logic that becomes hard to manage in drag and drop builders
- Little guidance on what is actually possible with agents
- A lot of technical setup, before anything feels useful
There is a huge drop off in the agent adoption funnel and we’re here to fix it.
The solution
Today we’re excited to launch the new version of Vellum.
Just explain the task you want to automate and Vellum turns it into a working agent (like Lovable for agents). Vellum will ask follow-up questions to understand what you need and connect to your tools. It handles the logic and makes the agent’s behavior visible, so you’re never guessing what it’s doing or why.
Agent: Getting from 0 to 1
Once you give Vellum your prompt, it asks a few follow up questions and builds a plan. For some use cases, it will ask which apps to connect and confirm access. When it has everything it needs, it creates the first version of the workflow:
Console: Access run and output details
After you have the first version of your agent, you get a UI that shows exactly how it works and what it does on each run. You can keep refining it with help from the agent itself, whether that’s improving prompts, adding new tools, or adjusting the logic.
To then test your changes, you can run the agent directly from the UI and inspect every decision it makes, like which tools it calls, when it decides to run a web search, and how it arrives at the final output.
We believe our agent logic is among the most reliable out there, and teams reach a working setup much faster compared to other agent builders.
Interact: Run manually, via triggers, API or code
Once you’re happy with how your agent works, you have a few ways to use it. Depending on your use case, you can choose one of the following:
- run the agent through a ready made UI
- vibe code your own UI with one of our integrations (e.g. Lovable)
- run it on a schedule or trigger it through APIs or app events
Most common use-cases
If you want to try Vellum today, here are the common agent automations we’ve seen ops teams build:
- Content generation: Agents around SEO and GEO keyword research, content analysis, and daily content creation.
- Sales ops automation: This is where we’ve seen the most use cases in the last few months. From automated demo prep, to pulling account intelligence from CRMs, to analyzing sales calls and deals.
- Product ops: Agents can reliably analyze support tickets and Linear issues, then ping the right DRIs to move things forward.
- Customer support: This is where agents really shine today. They can classify all incoming tickets and surface insights, like the best timing for renewals, automatically.
What we ask
- If you have one gnarly task that you’ve been putting out we ask you to login to Vellum and try to automate it. Share your feedback in this thread or text me at akash@vellum.ai
- Do any of these problems resonate with you? Please share your experience in creating agents to automate your tasks. We would love to learn more!
Vellum Workflows – Quickly prototype and deploy AI chains
Vellum - Build production-worthy LLM applications
How did your company get started? (i.e., How did the founders meet? How did you come up with the idea? How did you decide to be a founder?)
We worked together at Dover (YC S19) for 2+ years where we built production use-cases of LLMs. Noa and Sidd are MIT engineers who have worked DataRobot’s MLOps team and Quora’s ML Platform team respectively. Akash spent 5 years at McKinsey’s Silicon Valley Office. While working with GPT-3 and Cohere to build user-facing LLM apps, we found ourselves building complex internal tooling to compare models, fine-tune them, measure performance, and improve quality over time. This took away time from building our user facing product. We’ve worked on ML Ops for traditional ML and wished we had the same when later working with LLMs, so we’re building it.
