In a converted office a few blocks off University Avenue in Palo Alto, a small team is teaching software to do the part of your job you keep procrastinating. Not chatbots. Workers. The kind that finish things.
Who they are, right now
SuperAGI is a Palo Alto AI company building a platform of autonomous agents that handle real work end to end - prospecting and selling, writing code, running marketing campaigns, answering support tickets. It launched as a GitHub repo and an idea, raised $10 million from a quietly powerful investor, and now ships a suite of 25-plus AI-native apps that sit on top of a single agent runtime.
The bet, in one sentence - if software can write code, it can also do the meeting notes, the cold emails, the bug triage, the lead qualification, and most of the other things that fill a Monday. SuperAGI is racing to make that real before the rest of the market notices.
The work that nobody wants to do
Every company runs on busywork. Sales reps spend more time researching prospects than talking to them. Engineers spend more time tracking down a flaky test than writing the next feature. Marketers spend their Friday writing the email they meant to send Monday. Support teams answer the same five questions in a thousand polite variations.
The tools that promised to fix this - CRMs, ticketing systems, marketing clouds - mostly just gave the busywork a nicer interface. SuperAGI's founders looked at the stack and saw something different. The bottleneck wasn't the tooling. It was the assumption that a human had to be in every loop.
// One small irony - the workflow tools meant to save your week mostly added meetings about workflow tools.
Two operators, one hypothesis
Ishaan Bhola and Mukunda NS met at Navi, the fintech started by Flipkart co-founder Sachin Bansal in Bangalore. Bhola was head of product. Mukunda led engineering. They watched a generation of language models go from "interesting demo" to "actually useful" and concluded that the next layer up - agents that act, not just answer - would be where the real shift happened.
In 2023 they shipped the SuperAGI open-source framework. Developers starred it. Tens of thousands installed it. Then the founders did the unfashionable thing - they stopped chasing pure framework adoption and started turning the agents into products people would actually pay for.
- 2023 - SpringIshaan and Mukunda leave Navi to start SuperAGI.
- 2023 - SummerSuperAGI open-source agent framework released on GitHub.
- 2023 - LateKae Capital leads early backing. Headcount past 50.
- 2024 - JanuaryVEagle, an open-source vision model, released to the community.
- 2024 - FebruarySAM-7B small reasoning model published - beats Microsoft Orca-2 and GPT-3.5 on benchmarks.
- 2024 - March$10M Series A from Jan Koum's Newlands VC.
- 2024 - MidSuperSales and SuperCoder 2.0 launch as standalone commercial products.
- 2025+Platform expands to 25+ AI-native apps. Team grows to ~190 across four countries.
One platform, many workers
The SuperAGI platform reads less like a software suite and more like a roster. Each product is an agent that owns a function. They share memory, share context, share the same underlying runtime, and - notably - they get assigned tasks instead of being asked questions.
SuperSales
An AI SDR that researches, writes, sequences, and books meetings. Ships with a 250 million-contact lead database.
SuperCoder 2.0
Autonomous coding agent. Ranked #1 open-source on SWE-bench Lite at 34 percent.
SuperMarketing
Omnichannel campaign agent with segmentation, content, and analytics in one loop.
SuperSupport
An AI-native inbox that triages, drafts, and resolves the tickets that drain CS teams.
Open-Source Framework
The dev-first toolkit that started it all - still free, still on GitHub, still racking up stars.
SAM-7B / VEagle
Small, open reasoning and vision models published as research contributions.
// The platform's secret sauce - small models, big context, and the willingness to ship.
The numbers that change the argument
Claims are cheap in AI. Benchmarks are not. The SuperAGI team has chosen to compete on metrics where there is nowhere to hide - reasoning tasks for SAM-7B, real-world bug fixes for SuperCoder on SWE-bench. The results are the kind that make the rest of the open-source community pay attention.
The Series A came from Newlands VC, the quiet fund built by WhatsApp co-founder Jan Koum. Koum, an investor who rarely talks publicly about his bets, wrote the check in March 2024. That signal mattered more than the dollar amount. The dollar amount mattered too.
Agentic super-intelligence, working for a living
SuperAGI talks about "agentic super-intelligence" the way other companies talk about quarterly roadmaps. It is the founding bet expressed as a destination - an operating system for AI workers that runs the unglamorous parts of a business so the humans can do the rest.
The mission, stripped of the marketing - build agents that finish tasks. Make them cheap enough to deploy by the dozen. Make them open enough that developers extend them. Then put them inside the products people are already paying for.
Why this matters in 2026 and after
The agent market is loud and crowded. Most of it is wrappers around a foundation model and a Stripe checkout. SuperAGI's combination - a real open-source community, real benchmark numbers, a real commercial product, and a real team across four time zones - is the rarer thing.
If agents become the next layer of the software stack, the companies that own both the runtime and the apps will end up shaping how work feels. SuperAGI is one of a small handful of teams credibly playing for that position.
// What "agent platform" means depends on who is selling it. SuperAGI is shipping the version with the receipts.
Demos and interviews
Back to the converted office in Palo Alto
Walk past the office a year from now and the lights will likely still be on past midnight. But the work happening inside has already changed shape. The deck that used to take an analyst three days. The cold-email sequence that used to take an SDR a week. The pull request that used to take a backend engineer an afternoon. SuperAGI has been quietly handing those things to software that doesn't go home.
That is the punchline of the whole bet. Not that the agents will be smarter than the people. That they will be patient enough, and cheap enough, to do the parts the people would rather skip. The Palo Alto office isn't building an oracle. It's building a workforce.