The Top Agent-Pilled Startups in YC P26
YC's P26 batch goes to Demo Day next week, and one thing is clear: AI agents are no longer just writing code and summarizing documents. This batch is treating agents as first-class participants in the economy, complete with phone numbers, insurance policies, cloud infrastructure, and even the ability to hire humans. Here are the startups building around that thesis.
Thomas — The First AI Founder
Thomas is perhaps the most conceptually unusual company in the batch. Thomas is described as the first YC-backed AI founder: a virtual human who starts, runs, and grows his own companies on the internet. It's a bet that an AI entity can handle the full loop of company creation and operation, not as a tool for a human founder, but as the founder itself.
Humwork — Connect AI Agents to Human Experts in 30 Seconds
Humwork lets AI agents chat with human experts in real-time. When agents like Claude Code or OpenClaw hit a wall, Humwork's MCP server connects them to a verified domain expert in under 30 seconds, including senior engineers, lawyers, marketers, and designers. The founders, Yash and Rohan, compare it to Waymo's remote driver assistance model: AI handles most of the execution, but humans still sit at the edge for the hard decisions involving architecture, compliance, judgment, and tradeoffs. The two have been friends for 16 years, both attended UC Berkeley, and Yash is a 2x founder who built his first LLM startup in 2021.
RentAHuman — Marketplace for AI Agents to Hire Humans
RentAHuman flips the usual marketplace dynamic: instead of humans hiring AI, it's a marketplace where AI agents hire humans. When an agent encounters a task it can't complete autonomously, whether physical, regulatory, or requiring human judgment, it can source a human through RentAHuman to get it done.
AgentPhone — Phone Numbers for AI Agents
AgentPhone starts from a simple observation: agents can't act in the real world without a phone number, because they lack identity. The company gives every AI agent its own phone number to reach people and businesses through messages and calls. Developers get a single API to provision numbers with voice and messaging built in, handling compliance so they don't have to. The company was founded by brothers Manav and Meet Modi. Meet previously built AI agent infrastructure on WhatsApp for 280+ million businesses at Meta. Teams at Google's Agent Development Kit, Replit, Y Combinator, Sim AI, LangChain, and Alchemy are already building with AgentPhone.
Chronicle Labs — Staging Environments for Enterprise AI Agents
Chronicle Labs is a staging environment for enterprise AI agents: it captures every event an agent sees in production and backtests it, so customers can safely test new behaviors without breaking anything. As companies push agents into production workflows with real consequences, the ability to replay and validate agent behavior before deploying changes is increasingly critical.
InsForge — Agent-Native Cloud Infrastructure
InsForge (insforge.dev) is building a backend platform that AI coding agents can operate end-to-end. InsForge lets AI coding agents deploy and operate the full backend, including servers, databases, auth, and storage, from one consistent CLI. Instead of stitching together separate tools and dashboards, agents manage entire backend infrastructure end-to-end. The founders argue that coding agents are now the primary builders, but every backend platform was designed for human developers with dashboards, manual configs, and workflows that require a person at every step. The company was founded by Hang Huang, an ex-Amazon PM, and Tony Chang, an ex-Databricks network infrastructure engineer.
Klaimee — Liability Insurance for AI Agents
Klaimee is the first purpose-built E&O coverage for AI agents, covering what traditional E&O and cyber policies explicitly carve out. The timing is notable: major insurers like Berkshire, Chubb, and Travelers are winning state approval to exclude AI liability from standard commercial policies. Klaimee's underwriting process includes a public data scan, a 30-question governance assessment, and behavioral testing that fires more than 100 probes covering prompt injection, jailbreak, decision drift, and data leakage. Co-founder Ines Boutemadja was previously General Manager at SafetyWing (W18), where she spent five years building insurance products from scratch and scaled the startup from $5M to $60M.
Mount — The AI Insurance Carrier
Mount is taking a different angle on the same problem as Klaimee: it's building a full AI insurance carrier, starting with liability coverage for deployed AI agents. As companies move agents into production and give them more responsibility, Mount aims to be the carrier that helps businesses insure and secure them. Two P26 companies attacking AI agent insurance from different directions is itself a signal of how real the liability gap has become.
primitive — Communication for Agents
primitive is building the communication infrastructure needed for fully autonomous agents. As agents proliferate across organizations, they need ways to coordinate, hand off tasks, and share context with each other, not just with humans. primitive is positioning itself as that connective layer.
Standout — Agentic Hiring Marketplace
Standout gives every candidate and company an autonomous agent that represents them, assesses the other side, and makes introductions when there's a fit. Rather than a job board where humans browse listings, the idea is that agents on both sides of the hiring process handle discovery, evaluation, and matchmaking.
Superset — The IDE for the AI Agents Era
Superset (superset.sh) helps engineers run hundreds of coding agents in parallel. As the workflow shifts from writing code yourself to directing many agents simultaneously, current development environments weren't designed for that kind of orchestration. Superset is building the IDE that treats multi-agent coordination as the default.
YC P26 Demo Day is June 16, 2026.