AI Agents

Series A rounds are now the size of old growth rounds

The biggest recent agent rounds are labeled Series A but carry unicorn-scale numbers: $232M at $1.5B (Altimeter, Accel, Khosla, Felicis, 8VC), $220M at $1B (Lightspeed, Bessemer, Nvidia, Dell) and $160M at $1.5B. Series A is the most active stage in the last 90 days, with 33 deals and $7.3B. Tier-1 firms (Accel, Bessemer, General Catalyst, Sequoia at 6 deals each, plus a16z at 5) are crowding into the same rounds. Strategic capital from Nvidia and Dell Technologies Capital signals that infrastructure vendors want exposure to agent demand.

Recent launches focus on trust and control rather than new agents: BotGauge (red-teaming and evals for agents), Peras (spend limits on coding agents) and Mouse-1 (a cheap decision model for agent harnesses). Other launches include Notion Auto (agents running 24+ hours at lower cost) and Agent Hunt (rankings by real agent API calls). The company list also holds many evaluation and monitoring startups such as Judgment Labs, BentoLabs, Chronicle Labs and Prefactor. As agents run longer and spend real money, buyers need guardrails.

Why it matters · Picks-and-shovels tooling for agent QA, cost control and security is a less crowded bet than the agents themselves.

Several launches treat agents as teammates that need shared context and handoffs. BAND coordinates coding agents across tools, and Termi Protocol hands work between agents when rate limits hit. AgentGrid.io offers a shared human-agent workspace with version control. Local personal agents such as Bits and ShogunAI add human review or approval gates.

Why it matters · The orchestration and shared-context layer is where the next platform lock-in may form.

Composio Instant lets agents buy third-party tools without subscriptions or signups. ToDo.is runs always-on agents with audit trails, and Notion's agents run for 24+ hours. This points to machine-to-machine commerce and long-horizon autonomy becoming real product features. Existing companies in the data such as Paid and Nitro 4.0 sit in the same agent-billing and payments niche.

Why it matters · Agent-native billing and identity rails could become a new infrastructure market.

Basis is singled out for strong application-layer work in accounting, and Harvey and Legora are named as the legal AI leaders. Sierra is used by over 40% of the Fortune 50, and its Personas product lifted a telecom's resolution rate by nearly 50%. At the same time, commentators cite Harvey and Rogo as cautionary examples of AI labs moving up the stack into verticals.

Why it matters · Vertical agent startups need proprietary data and workflow depth to defend against labs moving up the stack.