AI Toolkits: The New Infrastructure Stack

About The Authors:

In our last blog, we broke down the architecture of an AI agent: the reasoning engine, the memory layer, the orchestration control plane, and the guardrails that keep everything on track. We described that, at its core, an agent is just a “brain in a jar”: able to think about actions but unable to execute them.  

Tools are how agents reach into the world: search the web, execute code, authenticate into third-party services, interact with browsers, and soon, operate machinery, place orders, and delegate to other agents. The model thinks. The tools act.

Just as the API economy created enormous enterprise value (Stripe, Twilio, and Plaid together are worth hundreds of billions), the tooling layer of the agentic stack will produce the same. But not all tools are created equal. Three categories are emerging, and each boasts unique risks and competitive dynamics.

Think of static tools as the library card of the agent era. They give agents access to information (the web, documents, and databases) without changing anything. These tools emerged first, are the most mature, and are already experiencing early-stage commoditization.

Take web search, for example. An LLM trained six months ago doesn’t know about last week’s earnings calls or yesterday’s regulatory rulings. Search APIs like Tavily, Parallel, and Exa bridge these gaps and provide structured, agent-optimized retrieval. Exa’s approach is particularly interesting, as the tool indexes the web semantically rather than by keyword.

Then there is the problem of unstructured data (PDFs, Slack archives, internal wikis, scanned contracts, etc.). Morphik and Unstructured (AV portco) specialize in converting this messy reality into clean, retrievable context that agents can actually use. Firecrawl and Apify, in particular, specialize in structured web extraction and turn dynamic pages into clean, structured data that agents can process. As agents are increasingly tasked with market research, competitive intelligence, and price monitoring, this category stands to dramatically increase in scale.

Dynamic tools are how agents go from observation to action. They’re also how agents become risky. A tool that can search the web is safe. A tool that can execute code, authenticate into production systems, and interact with browsers on a user’s behalf is a different beast entirely.

When it comes to investing in the AI agent stack, most VC dollars are flowing here and for good reason. The problems are hard, the moats are real, and enterprise demand is obvious.

Holding this dynamic ecosystem together is Model Context Protocol (MCP), Anthropic’s open standard for how agents communicate with tools. It is fast becoming the mechanism that lets any agent talk to any tool without custom integration. Companies building on MCP today have a distribution advantage that will compound as the standard proliferates.

Static tools let agents know. Dynamic tools let agents do. Emerging tools let agents exist across agentic networks and sensory modalities.

This tooling layer is the frontier and, by definition, early, messy, and high-risk. But it is also where we expect to see the largest VC returns in the coming decade.

Looking Ahead

As you can see, the tooling layer of the agentic era is still being built. Most of the companies discussed did not exist three years ago. The category did not even have a name two years ago. We are in the early innings of this buildout. And the tools that will win will be those closest to production-critical workflows, hardest to replicate, and compounding with agent adoption.

This communication is from Alumni Ventures, a for-profit venture capital company that is not affiliated with or endorsed by any school. It is not personalized advice, and AV only provides advice to its client funds. This communication is neither an offer to sell, nor a solicitation of an offer to purchase, any security. Such offers are made only pursuant to the formal offering documents for the fund(s) concerned, and describe significant risks and other material information that should be carefully considered before investing. For additional information, please see here. Achievement of investment objectives, including any amount of investment return, cannot be guaranteed. Co-investors are shown for illustrative purposes only, do not reflect all organizations with which AV co-invests, and do not necessarily indicate future co-investors. Example portfolio companies shown are not available to future investors, except potentially in the case of follow-on investments. Venture capital investing involves substantial risk, including risk of loss of all capital invested. Diversification cannot prevent investment loss; it is a strategy to mitigate investment risk. This communication includes forward-looking statements, generally consisting of any statement pertaining to any issue other than historical fact, including without limitation predictions, financial projections, the anticipated results of the execution of any plan or strategy, the expectation or belief of the speaker, or other events or circumstances to exist in the future. Forward-looking statements are not representations of actual fact, depend on certain assumptions that may not be realized, and are not guaranteed to occur. Any forward-looking statements included in this communication speak only as of the date of the communication. AV and its affiliates disclaim any obligation to update, amend, or alter such forward-looking statements, whether due to subsequent events, new information, or otherwise.

Interested in Seeing Elite Venture Deals (for Free)?