Building the Open Runtime for AI agents.
Agents are becoming real parts of real software, and they multiply. Nasiko is the layer that keeps them in check: what each one costs, what it can touch, and what it does at runtime, from a developer’s laptop to production.
Open source, and works with any model, agent framework and cloud. Nasiko is a venture-funded Silicon Valley company with roots in MIT’s AI agent research.
The people behind Nasiko
10+ years of research, 20+ papers, and 15+ university and industry research partners behind the platform.
-
Rahul Todkar
Co-founder & CEO
Visionary AI leader with 20+ years leading and scaling AI products for 300M+ users and more. Previous Head of Data and AI at Tripadvisor, leadership roles at LinkedIn, Intuit, PayPal and Capital One. Track record of turning emerging technologies and early concepts from pilots to enterprise scale adoption with huge business ROI.
-
Karan Bharadwaj
Co-founder & Head of Product
A decade building across AI, Web3 and fintech. Previously CTO at XDC Protocol and Mool, founder of Arithmic Labs, and technical advisor to Project NANDA.
-
Ramesh Raskar
Co-founder & Chief Scientist
Associate Professor at MIT. Former innovation lead at Facebook, and led special research projects at Google and Apple. Holds 125+ patents and has won the Lemelson, ACM SIGGRAPH and Academy of Inventors awards.
Backed and recognized by
Enterprise AI is moving from individual agents to agent swarms, but legacy infrastructure presents a false choice: constrain agents until they lose their value, or give them autonomy without accountability. Nasiko, an MIT spinout, is the open, extensible runtime that removes that tradeoff. Starting with TokenOps, it gives enterprises the flexible infrastructure to scale agent capability with the security, performance, and cost controls required for production.
Vijay ReddyPartner, Mayfield
Frequently asked questions
Where is Nasiko based?
Silicon Valley, with roots in MIT’s AI agent research.
Why “Open Runtime”?
Nasiko is open source, so you can read the code and run it yourself. It’s also open by design: use any model, any agent framework and any cloud, and switch any of them without rewriting your agents. No vendor locks you in, including us.
What problem is Nasiko solving?
Agents are becoming real parts of real software, and they multiply. What’s missing is the layer that makes them work together and keeps them in check: what each one costs, what it can touch, and what it does at runtime. That’s the layer Nasiko builds, from a developer’s laptop to production.
Is Nasiko research-led?
Yes. The team publishes the ideas the platform is built on, including PACMS, a method for choosing which context an agent sends to the model. It also writes openly about agent security, memory, routing and token costs. Our research page lists the papers.
How seriously does Nasiko take security?
We built security in before we needed it. Agents never hold credentials, every tool call is checked against policy, and our software supply chain is hardened. You can report vulnerabilities privately through GitHub or [email protected]: we acknowledge within 2 business days and share a triage verdict within 7. SOC 2 Type 2 and ISO 27001 are in progress.
How do I get started?
Get the open-source code on GitHub, or start free in the web app at nasiko.com, no card needed. You can also install the desktop app or deploy from the DigitalOcean Marketplace. Want a walkthrough? Book a demo.



