Lightsage, a San Francisco startup building infrastructure for the age of autonomous AI agents, announced September 8 it closed a $4 million seed round led by Nexus Venture Partners. The financing included participation from Steven Tamm, former chief technology officer at Salesforce, alongside Abhinav Asthana, chief executive of Postman, Matt Curl, chief executive of Apollo, Robert Chatwani, president and general manager of Growth at DocuSign, and several founders from the developer tools ecosystem.
The company, which operates under the legal name API-Rex Inc., is placing a bet on a shift that still feels speculative to many in enterprise software: that artificial intelligence agents will soon do more than suggest code or answer customer service questions. They'll evaluate products, integrate APIs, and negotiate purchases without meaningful human oversight.
Lightsage's platform addresses a technical challenge that emerges from that premise. When coding agents like Claude Code, Cursor, or GitHub Copilot attempt to use a software company's API, they often stumble over documentation that made sense to human developers but confuses machine interpretation. Authentication protocols trip them up. Design choices that were intuitive in one context become opaque in another.
The startup has built what it describes as a three-part system. Benchmark evaluations test whether agents can actually complete tasks using a given API. Diagnostics isolate where the friction occurs, whether in documentation, authentication flows, or underlying design. Analytics track how agents interact with products in real time, surfacing patterns that might otherwise go unnoticed.
Early customers include Firecrawl, Reducto, Daytona, Rime, and TinyFish, all companies already selling tools to developers who work extensively with AI systems. That customer base is narrow by design. Developer infrastructure represents the leading edge of agent adoption, though whether the broader software market follows remains an open question.
From benchmark experiments to commercial infrastructure
Before rebranding as Lightsage earlier this year, the company operated as Sapient and ran DevTool Arena, a public leaderboard that evaluated how well more than 140 developer-focused companies performed in agent-driven scenarios. According to a LinkedIn post from August by CEO Jun Liang Lee, the benchmark conducted over 3,000 evaluations. Companies tested included ElevenLabs, Stripe, Browserbase, Pinecone, Twilio, Resend, and Clerk.
Lee and co-founder Sean (Wen Han) Er spent months organizing what they called Agent Experience Demo Nights across San Francisco, gathering feedback from product teams at Auth0, AWS, and Okta. Those sessions appear to have informed the pivot from open benchmarking to a commercial product sold to companies anxious about how agents perceive their software.
The repositioning to Lightsage came ahead of the seed close, though the company has declined to share revenue figures or specifics on user traction. According to Forbes, some customers claim to have seen sales to AI agent users increase by 10 to 20 percent, citing Lee, though those figures have not been independently verified.
Nexus Venture Partners, which manages $3.2 billion and has backed Apollo.io, Postman, and Firecrawl, led the investment. Abhishek Sharma, a partner at the firm, framed the opportunity as "Agent-Led Growth," positioning it as the natural evolution of product-led growth and developer experience strategies that dominated the last decade of enterprise software.
Software marketing enters uncharted territory
The underlying thesis echoes a comparison industry observers have started making: AI agents discovering and purchasing software may reshape markets the way search engines once did. Companies that optimized for Google's algorithms gained structural advantages. Now, the argument goes, businesses need to optimize for how Claude or GPT-4 interpret API documentation.
Analysts have floated predictions that sound dramatic even by the standards of technology forecasting. Forrester suggested that by this year, a fifth of business-to-business companies might face negotiations conducted entirely by AI agents, according to Forbes. Forbes also reported a Gartner projection that by 2028, 90 percent of B2B purchasing could involve AI intermediaries, representing some $15 trillion in transactions. Neither forecast was linked to original research, and such projections deserve the usual skepticism reserved for long-range technology predictions.
Still, capital has begun flowing into adjacent infrastructure. Firecrawl raised $14.5 million in August 2025 from Nexus. Daytona and Rime each closed $24 million Series A rounds more recently. Arena.ai, connected to the LMSYS research ecosystem, launched agent benchmarking tools that overlap conceptually with what Lightsage built in its earlier form.
The pattern suggests investors see enough early signal to place bets, even if the full vision of autonomous agent commerce remains years away from mainstream adoption.
"We are moving from an internet where AI tells people which software to use to one where AI increasingly uses the software itself," Lee said in the announcement. "That changes what growth means. Visibility still matters, but the real test is whether an agent can understand your product and get to a successful outcome."
Lightsage plans to expand beyond coding agents into other developer surfaces like SDKs, command-line interfaces, and MCP servers, then eventually into broader B2B categories including infrastructure and payments software. The company has posted roles for a founding go-to-market hire, signaling plans to build out commercial functions alongside its technical team.
Whether that expansion happens on the timeline investors expect may depend less on Lightsage's execution than on how quickly agents move from experimental tools to primary enterprise buyers. For now, the company appears to be building for a future that hasn't quite arrived but that enough people find plausible enough to fund.
