By now, we have all seen agents in ChatGPT and Gemini that are specifically about researching. However, Superagent is built differently – at least that’s what Airtable CEO Howie Liu suggests.
Calling Superagent “just another AI assistant” is like calling a Tesla “just another electric scooter.” Airtable’s latest brainchild doesn’t just add AI on top of your workflow, it aims to rethink what a workflow is when intelligence is embedded at the core.
Here’s where Superagent aims to be different. Whether it’s ChatGPT, Grok or Gemini, we have all grown used to the solo-AI experience, where we ask a question and get a well-mannered paragraph in return.
Superagent does things differently, more in-depth. Ask it a question, and what you get is the output of a team – or at least, what feels like one. Think charts, timelines, strategy decks. It doesn’t summarize, it creates a briefing document, claims CEO Howie Liu.
Also read: Want to build AI agents? 5 simple ways to start for beginners
At its core, Superagent claims to be a multi-agent system. Not one AI doing everything, but a small army of specialist AIs – a science analyst, a market researcher, a product strategist, etc – each activated in parallel. They’re all essentially breaking down the scope of your prompt to solve a problem or execute a complex task, then stitch all of its research back into a single, presentation-ready response.
“You’re not prompting an AI,” Liu explains, in an interview with Techcrunch. “You’re orchestrating a team.” In other words, it isn’t a glorified Zapier flow with an LLM stuck on top – it’s a full-fledged planning system. Superagent coordinates, deploys work, adapts in real-time, and surfaces ideas you didn’t even thought of asking in the first place.
That last bit – uncovering unprompted insight – might be the killer feature, if true. Ask about market expansion, and Superagent won’t just name cities. It’ll show you heatmaps, demographic breakdowns, risk factors, and even surface a niche regulation that might trip up your product team.
Also read: Cursor AI agents just wrote 1 million lines of code to build a web browser from scratch, here’s how
Airtable CEO Howie Liu’s confident in where Superagent sits in the emerging AI agent landscape. He nods respectfully to Anthropic’s Claude and a lesser-known peer, Manus, for their architectural rigour — calling them “generally capable, long-running and really smart agent architecture.”
According to Liu, everything else that claims to be an AI agent or chatbots trying to offer agentic workflows are nothing but LLMs doing all the heavy lifting. They all have predetermined steps with simple AI look up calls preconfigured, which is a long way away from true autonomous agents that can do nuanced search and retrieval of information or backtrack their course when presented with new learnings in real-time.
So in the eyes of CEO Howie Liu, Superagent’s only true competition in the market is Claude and Manus (which Meta is trying to acquire). This sums up Liu’s belief that Superagent isn’t just an experimental leap, but a true platform shift. It isn’t generative wallpaper. It’s operational intelligence like never presented before by an AI tool.
In terms of pricing at launch, Superagent will run on a tiered model – just like ChatGPT and others. A standard $20/month for individual Pro users, and up to $200/month for power users, with built-in inference credits.
Also read: AI agents can’t do complex tasks, claims ex-Infosys CEO
