--- title: AI Growth Startup Lightsage Raises $4M to Help Software Companies Turn AI Coding Agents Into Customers description: Nexus Venture Partners leads a $4M seed for Lightsage, whose platform tests how AI coding agents like Claude Code and Codex discover and use software. author: Darie Nani (Editor-in-Chief) date: 2026-09-08T13:53:25.670Z updated: 2026-09-08T13:53:25.689Z canonical: https://www.sovereignmagazine.com/article/lightsage-4m-seed-agent-led-growth-nexus image: https://cdn.nanimediahouse.com/lightsage-founders.webp categories: Artificial Intelligence, Startups content_type: Spotlight region: San Francisco publication: Sovereign Magazine about: - type: Organization name: Lightsage description: Lightsage is a San Francisco software company that helps developer tools and B2B software companies understand and improve how AI coding agents such as Claude Code, Codex and Cursor discover, use and pay for their products. Its Agent-Led Growth platform runs simulated agent tasks across APIs, SDKs, CLIs and MCP servers, pinpoints where agents fail, and tracks real agent traffic on a company's website and documentation. Founded by CEO Jun Liang Lee and CTO Sean Er, Lightsage raised a $4 million seed round led by Nexus Venture Partners in September 2026. url: https://lightsage.com foundingDate: 2024-01-01T00:00:00.000Z industry: AI software --- Lightsage, a San Francisco startup that measures how [AI coding agents](https://www.sovereignmagazine.com/article/cielara-code-ai-coding-localization) find, install and use software products, has raised $4 million in seed funding led by Nexus Venture Partners. The round also includes angel investments from former Salesforce chief technology officer Steven Tamm, Postman chief executive Abhinav Asthana, Apollo.io chief executive Matt Curl, DocuSign president Robert Chatwani, Resend chief executive Zeno Rocha, Firecrawl co-founder Eric Ciarla and Daytona chief executive Ivan Burazin. The company sells what it calls an Agent-Led Growth platform. Software companies use it to see how agents such as Claude Code, [Codex](https://www.sovereignmagazine.com/article/openai-codex-plugins-ai-coding-tools), Cursor and GitHub Copilot behave when given a task that involves their product, from reading the documentation to authenticating and calling the API. When an agent gives up or picks a competitor, Lightsage records the point of failure so the vendor can fix it and run the task again. ## Product-Led Growth and the Nexus Venture Partners Portfolio > "AI is now shifting that agency from humans to agents, which can discover, evaluate and act on a customer's behalf. Lightsage is building the intelligence infrastructure for this new era of autonomous browsing, helping companies optimize for agent conversion, not just awareness." > — Abhishek Sharma, Nexus Venture Partners Nexus, which closed its eighth fund at $700 million in December 2025, has backed product-led growth companies for a decade. It wrote Postman's first $1 million seed check in 2015, led Apollo.io's $7 million Series A in 2018, and led Firecrawl's $14.5 million Series A in August 2025. All three grew by letting individual developers or salespeople try the product free and pay later, with no sales call in between. Abhishek Sharma, the Nexus managing director who led the Lightsage investment, has spent most of his time at the firm on companies with that kind of bottom-up sales motion. Product-led growth assumes a person is on the other end. Someone searches, lands on a website, signs up, reads the docs and either becomes a paying customer or leaves. Every tool in the marketing stack, from analytics to onboarding emails, is designed to measure and improve that path. Sharma's argument for Lightsage is that the same path now increasingly has an AI agent on it, and none of the existing tools can see what the agent does. ## AI Coding Agents as the New Software Buyer > "We are moving from an internet where AI tells people which software to use to one where AI increasingly uses the software itself." > — Jun Liang Lee, Lightsage A coding agent asked to add email sending, a database or a payment flow to an application can now search for options, choose a provider, install its SDK and write the integration without the developer visiting a single vendor website. Stack Overflow's 2025 developer survey found that 23% of developers already used AI agents regularly in their work, and Anthropic said in February that Claude Code alone had passed a $2.5 billion annual revenue run rate. Lightsage's position is that visibility to an agent is only the first step. An agent that finds a product still has to understand the documentation, pick the right SDK, get through authentication and reach a working result. If it fails at any of those points, it moves on to a competitor, and the vendor never learns that a sale was lost. "Visibility still matters, but the real test is whether an agent can understand your product and get to a successful outcome," said Jun Liang Lee, Lightsage's chief executive and co-founder. "We think every software company will eventually need to optimize for that journey to unlock Agent-Led Growth." The phrase has been in use among investors since Insight Partners published a note on it in March 2026. That note cited data showing Claude Code chose Resend for email sending 63% of the time, against 7% for the much larger SendGrid. Resend's chief executive is one of the angels in the Lightsage round. ## Agent Experience Testing Across APIs, SDKs, CLIs and MCP Servers Lightsage runs large-scale simulations across coding agents and answer engines. Agents are given real tasks that require them to read documentation, choose tooling and use a product through its API, SDK, command-line tool, MCP server or agent skills. The platform measures where a company appears against competitors and what happens after that, and when an agent fails it identifies the cause, which might be discoverability, confusing docs, authentication, a broken endpoint or an incompatible MCP server. The company published its own benchmark in April. It ran more than 70 APIs through Claude Code and Codex, spending more than $5,000 in agent usage, and found that the time an agent needed to get a product working varied by a factor of 50 between the fastest and slowest. Fewer than half of the APIs tested offered a command-line tool. Lightsage's data also suggests that typed SDKs and a working MCP server add noticeably to an agent's success rate, though those figures are the company's own and have not been independently checked. A second part of the platform tracks real agent traffic on a company's website and documentation, showing what agents look at and whether those visits turn into product usage. Mintlify, which hosts documentation for thousands of developer tools, said in April that AI agents now account for about two thirds of requests to the docs it serves, so the traffic is already there to measure. The platform currently supports Claude Code, Codex, Cursor, GitHub Copilot, OpenCode and [other coding agents](https://www.sovereignmagazine.com/article/testmu-kane-cli-launch). Pricing starts at $250 a month for startups. ## Generative Engine Optimization and AI Visibility Tools The closest established category is generative engine optimization, or GEO, which measures how often a brand is mentioned in answers from ChatGPT, Perplexity and Google's AI Overviews. Profound raised $96 million in February at a $1 billion valuation, and Peec AI raised $21 million last November. Those tools measure whether a brand is mentioned, and Lightsage's argument is that for software companies a mention is not a sale, because the agent still has to install and use the product. Its platform picks up after the mention and follows the agent through to a working integration. On the traffic side, Vercel said in February that it had built an internal tracker for how coding agents search the web after finding they did so in about a fifth of prompts, which suggests larger companies are starting to want the same data. ## Early Customers and the Angel Investors Lightsage is starting with developer software because agent behavior there is easy to observe. Customers include Firecrawl, Reducto, Daytona, Rime and Tinyfish, which use it to understand why agents choose one product over another, where integrations break and how key workflows hold up after a product or documentation change. A typical engagement starts with a coding agent repeatedly recommending a competitor. Lightsage reproduces the same task across products and agents to isolate the cause. Several of the angels run companies that sell to developers and depend on being chosen by their agents. Asthana's Postman is the most widely used API platform. Curl's Apollo.io sells sales software. Ciarla's Firecrawl and Burazin's Daytona are both Lightsage customers as well as investors. Other angels include Hila Qu, formerly head of growth at GitLab, Tinyfish chief operating officer Shuhao Zhang and Adam Frankl, who was the first marketing lead at JFrog, Neo4j and Sourcegraph. ## Lightsage Founders and Plans for the Funding Lee and chief technology officer Sean Er founded the company after Er spent three months at ByteDance writing internal documentation. The pair went through the Alchemist Accelerator in San Francisco, where Frankl is a partner, and earlier sold a tool that generated sample applications for API companies. Lee, who studied at Stanford and the National University of Singapore, also runs a recurring AI developer tools demo night in San Francisco that has drawn more than 3,000 attendees across four editions. The funding will go toward agent evaluation, analytics, attribution and optimization across APIs, SDKs, command-line tools, MCP servers and agent skills. Developer tools are the first market, and as agents begin acting directly across business software, infrastructure and payments, Lightsage wants to be the system companies use to understand and improve how they perform in that channel. It is hiring across technical and commercial roles. ## FAQ **Q: What is product-led growth?** A sales model in which the product itself brings in and converts customers. Users try the software free or on a low tier, see the value and upgrade, with little or no involvement from a sales team. Postman and Apollo.io, both in the Nexus portfolio, grew this way. **Q: What is agent-led growth?** Lightsage's term for the same idea when the user is an AI agent rather than a person. A coding agent finds a product, reads its docs, installs its SDK and starts using it, and the vendor's job is to make sure the agent gets to a working result. Insight Partners used the same phrase in a March 2026 note. **Q: How does an AI agent decide which tool or API to use?** Agents such as Claude Code and Codex combine what their underlying model already knows with web search and documentation lookups, then pick the option they can get working with the least friction. Lightsage's April benchmark found that clear docs, typed SDKs, a command-line tool and a working MCP server all raised the chance of an agent succeeding with a given API. **Q: What does Lightsage's platform actually do?** It runs simulated tasks across coding agents and answer engines to see whether an agent can discover, understand and successfully use a company's product, reports where and why it fails, and tracks real agent traffic on the company's site and documentation. **About Lightsage** Lightsage is a San Francisco software company that helps developer tools and B2B software companies understand and improve how AI coding agents such as Claude Code, Codex and Cursor discover, use and pay for their products. Its Agent-Led Growth platform runs simulated agent tasks across APIs, SDKs, CLIs and MCP servers, pinpoints where agents fail, and tracks real agent traffic on a company's website and documentation. Founded by CEO Jun Liang Lee and CTO Sean Er, Lightsage raised a $4 million seed round led by Nexus Venture Partners in September 2026. [Website](https://lightsage.com)