Market

A $100B+ market is being built for non-human customers.

The agent economy is roughly where e-commerce was in 1998. The infrastructure is missing, the categories are open, and the customer count is about to 100x.

10 yrs

TAM ramp for agent-native infra — comparable to mobile 2008–2018

10×

agent customers vs human SaaS buyers by 2030

24/7

agent purchase cycles — no demos, no sales calls, no business hours

Ten categories getting rebuilt

Each one has incumbents that were designed for human buyers and human integrators. None of them are agent-native. The new winners look very different.

• Payments¤

  Charge cards, send invoices, move funds between agent wallets.

• Identity & KYC◇

  Prove humanness, run KYC, issue and verify decentralized identifiers.

• Memory◉

  Long-term, shared memory and semantic recall for agents and their teams.

• Browser actions⌘

  Headless browsing, structured clicks, extraction, and captcha fallbacks.

• Comms (A2A)↯

  Agent-to-agent messaging, negotiation, and signed handoffs.

• Voice & phone☏

  Place and receive calls, transcribe, and route to humans when needed.

• Research✦

  Web search, summarization, and citations with freshness guarantees.

• Code executionλ

  Sandboxed runtimes for code, data, and tool-use loops.

• Real-time data≋

  Streaming prices, weather, traffic, sports, and other live feeds.

• Trust & safety✓

  Provenance, abuse signals, and signed attestations for agent behavior.

Why now

• Frontier models can hold tool-use loops for long enough to be customers, not just toys.
• Stablecoin and agent-wallet rails make per-call billing real.
• MCP, A2A, and similar protocols normalize the interface between agents and tools.
• Every major platform is shipping an agent runtime. The aggregate volume is already huge.

Comparable shifts

1995 web. 2008 mobile. 2014 cloud. Each one looked janky for a year, then obvious. Agent-native is the next one. Agent Oasis is the registry layer for it.