A new Agentic AI startup, Skyfire, recently emerged from stealth with $8.5 million in seed funding to enable AI Agents to autonomously spend and manage money.
“Our payment layer enables AI Agents (and multi-Agent systems) to autonomously make and receive payments, and hold balances,” said Skyfire’s CEO and co-founder, Amir Sarhangi, in an interview with StartupHub.ai. “Think of us as the fintech infrastructure for Agentic AI.”
Over the past year, an ecosystem of Agentic AI has rapidly taken shape, spanning from developer tools to consumer-facing applications. Frameworks like LangChain, Crew AI, and Composio empower developers to build AI Agent systems and workflows capable of executing complex tasks through API integrations and workflow management. These advancements have sowed a burgeoning market segment that Sarhangi describes as the “most active startup pocket.”
Microsoft’s introduction of AutoGen Studio, a low-code interface to prototype AI Agents, packing another punch in their Agentic AI framework, AutoGen.
The subsector is aptly characterized as ultra nascent, but it has the makings to reach a level of maturity quickly. Large Language Model (LLM) developers are adapting their capabilities to handle the large context window needed to orchestrate long workflows. And testing and observability capabilities are forming to ensure LLM apps and AI Agents alike function as intended.
Despite this rapid progress, inherent shortcomings persist. Function calling, designed to enable LLMs to interact dynamically with external tools and APIs, has shown promise but also faced challenges with reliability, leading to inconsistent execution. Similarly, OpenAI’s Structured Outputs, intended to ensure AI-generated data conforms to specific schemas, have encountered issues such as slower response times and occasional errant schema adherence. While these features mark steps forward, they remain unpropitious, lacking the ability to autonomously handle financial transactions—part and parcel with our expectations of autonomous AI Agents.
This is where Skyfire adds timely value.
Skyfire’s secure payment system allows users to provide AI Agents with a set amount of money, enabling them to spend autonomously—akin to giving your AI assistant a company credit card, but with controls to ensure responsible spending.
According to Sarhangi, AI Agents are expected to conduct $46 billion in commerce by 2027, encompassing transactions with vendor APIs and between AI Agents themselves, establishing a new business model called Agent-to-Agent commerce (A2A).
Some industry figures, like Box CEO Aaron Levie, have even stronger expectations for the market’s potential, highlighting AI Agents as a $1 trillion opportunity.
Box CEO Aaron Levie on why AI agents are a $1 trillion opportunity
“AI will not just be serving back an answer for you, it will also go and do something for you … as tokens come down in cost, we can scale up intellectual labor and knowledge work [with AI] orders of magnitude… pic.twitter.com/Pw5Bm4JNqI
— Overlap: Business & Tech (@Overlap_Tech) August 30, 2024
Skyfire’s platform is designed as a comprehensive financial stack for the AI economy, providing essential tools and protocols for AI Agents to operate independently. Their open, global payments protocol enables AI Agents to access LLMs, datasets, and API services without needing traditional payment methods, ensuring seamless global transactions.
The platform also supports automated budgets and control, allowing developers to set spending limits, ensuring AI Agents operate within predefined business parameters. Security is reinforced through Agent identification and history verification, which ensures authentication and authorization. “We can verify whether an Agent is owned by the person it claims to be and if it’s operating within the parameters set by its owner,” explained Craig DeWitt, Skyfire’s co-founder and Head of Product. “There will be individuals and businesses that will only transact with others who have gone through that verification process.”
AI Agents can be funded through traditional banking methods or stablecoins, with transactions completed instantly. Skyfire also supports fractionalized payments, allowing for pay-as-you-go models where AI Agents purchase only what is needed based on usage and pricing parameters.
Skyfire initially targets LLM developers and AI Agent builders. “There are over 160 LLMs today,” Sarhangi pointed out. “We enable developers to use any of these LLMs through our protocol without needing to open separate bank accounts or hold balances at each one.”
While the idea of AI Agents handling real money might be a point of concern, Skyfire’s founders are confident in the system’s security. DeWitt emphasized that the platform is designed to be as secure as any traditional online payment system. “The only thing required for sign-up today is an email address,” said DeWitt. “From that, users can fund accounts, set balances, and connect to service providers through Skyfire.”
Together, Sarhangi and DeWitt have built scalable, global software and payments infrastructure. Sarhangi, formerly VP of Product at Ripple, played a pivotal role in developing instant, universally accepted payment solutions. DeWitt, also a former Ripple executive, contributed to foundational payments technology integral to the cryptocurrency and blockchain industry.
Skyfire’s payment network is now open to Agentic AI developers, LLMs, and API providers for integration.
With digital commerce expected to reach $9.8 trillion by 2027, Skyfire’s technology is poised to drive the growth of A2A commerce, and even market expansion. Gartner echoes this projection, forecasting that AI Agents will account for a significant 33% share of digital interactions within the Generative AI market over a similar time frame.
The startup’s seed round was backed by Neuberger Berman, Brevan Howard Digital, Intersection Growth Partners, DRW Venture Capital, Inception Capital, Arrington Capital, Red Beard Ventures, Sfermion, Circle, FBG, Gemini, Crossbeam Venture Partners, EveryRealm, Draper Associates, ARCA, and Ripple Labs.
