Robinhood Ventures Fund II is opening a door for retail investors into the early-stage AI economy. The fund holds roughly 80 positions built on $250K SAFEs, leans heavily on Y Combinator alumni, and begins trading on the NYSE on August 13 under the ticker RVII with shares priced at $25. A significant slice of the portfolio is devoted to agentic AI, the startups building software that plans, acts, and works alongside humans instead of just answering questions.
The agentic AI positions split into distinct layers of the stack. One cluster is solving trust. Agentic Fabriq builds identity and permissioning for AI agents, giving each autonomous worker a verifiable identity and a bounded set of actions it is allowed to take. As agents start spending money, editing documents, and contacting customers, that control layer becomes the difference between a useful assistant and a liability. The company is a reminder that the agent economy will need more than capable models; it will need rules about who agents are and what they may do.
Another group is working on how agents show up in human settings. Keyframe Labs creates lifelike AI avatars for video calls, letting an agent appear on screen with realistic expressions, gestures, and movement instead of a text box. For sales, support, and onboarding workflows, the avatar is the interface, and making it believable is the product. The bet is that people will accept AI participants in meetings far more readily when they look and behave like participants, and that the companies building that presence layer will capture real value as video becomes the default way businesses talk to machines.
Communication between agents is the third layer. Second Stage Labs, which builds under the Primitive brand, is attacking agent-to-agent communication with infrastructure that lets one agent discover, message, and coordinate with another. That plumbing is a prerequisite for fleets of agents working a single task, and it points at a future where most AI-to-AI traffic never involves a human at all. For the fund, the position is a way to own a piece of the network layer that every connected agent will eventually use.
Safety and testing are a recurring theme. Arga Labs builds sandboxes where developers can test agents against realistic scenarios before release, catching broken workflows, hallucinated steps, and runaway behavior before they reach production. As autonomous systems take on higher-stakes work, the tooling to fail safely in a test environment becomes as important as the agent itself. A company that ships an unreliable agent once can lose customer trust permanently, which is why the testing layer is quietly becoming a standard part of the agent stack.
Scheduling is a deceptively hard problem, and Apex Flux applies agents to complex scheduling that spans people, calendars, and shifting constraints. It is one of several bets on agents taking over the coordination work that currently consumes human hours in every large organization. Scheduling sits at the intersection of reasoning and real-world logistics, exactly the kind of messy, constraint-heavy problem where early agent products can prove their value.
AutoSitu, which also goes by MirageDoodle, builds an AI-native workspace for development plan reviews. The tool helps engineering teams walk through technical plans with an assistant that reads, questions, and annotates the proposal, moving review conversations out of scattered comment threads and into a structured AI-assisted workflow. For teams shipping software on tight cycles, that compression of the review process can translate directly into faster delivery.
Veriad rounds out the productivity angle with an AI email and calendar assistant that manages inboxes and schedules under user control. The company targets the same inbox friction most knowledge workers feel daily, automating the triage while keeping a human in the loop on anything consequential. Where other assistants generate drafts, Veriad’s focus on permissioned, user-controlled action is a bet on agents that can be trusted to follow through.
Taken together, the agentic AI positions read as a bet on the software layer that will run above foundation models. Identity, communication, testing, and interfaces are all pieces of an agent stack in the making, and the fund is spreading its chips across every level of it rather than picking a single winner. That structure suits a fund that cannot manage dozens of individual startups closely; it lets the portfolio benefit from the category as a whole.
Robinhood priced the fund at $25 a share ahead of its August 13 NYSE debut under the RVII ticker. The $250K SAFE structure and the Y Combinator lean give the portfolio an early-stage profile, and the agentic AI holdings signal where the fund’s managers expect the next wave of value to be created. For retail investors, the fund is a way to hold a diversified slice of the agent economy without picking individual winners.
