Every serious AI workflow in 2026 has the same unsolved problem: you can measure what the AI did, but you cannot settle it in real time at token granularity. Not on Stripe. Not on Ethereum. Not on any cheap L2.
P402 Meter is the proof that the problem is solved. Four use cases, one settlement layer, sub-millidollar economics. Every AI token priced and settled onchain as it happens.
It is not a product gap. It is a structural constraint. The economics of settlement on every existing rail make per-action AI billing irrational before you write a line of code.
A 50-token AI response costs $0.00003. Stripe's minimum is 577× that. You'd need to batch 10,000+ micro-actions before the rail makes economic sense — destroying real-time attribution.
structurally broken for AISettlement costs 95,000× the AI work itself at $0.00003/response. Every per-token settlement burns more in gas than the revenue it records. Non-starter at any volume.
economically irrationalBetter, but still 300–3,000× the cost of the AI action. At $0.01/settlement, a 55-event session costs $0.55 in gas alone — 1,600× the actual AI compute cost.
still 2–3 orders of magnitude offTempo's FeeAMM settles gas in TIP-20 stablecoins (USDC.e), not a native token. There is no ETH, no gas wars, no minimum viable transaction size imposed by the protocol. MPP (Machine Payment Protocol) provides the programmable payment orchestration layer on top.
Tempo mainnet (Chain ID 4217). USDC.e TIP-20 transfers via FeeAMM. Gas paid in stablecoins. <$0.000001 per settlement. One real Tempo tx per session aggregates all provisional events into a verifiable onchain proof.
Machine Payment Protocol. Pre-funded TEMPO_TREASURY_PRIVATE_KEY wallet. Per-session-close settlement pattern: 55+ provisional ledger events, one real USDC.e transfer at stream close. mppx dual-402 challenge/response for machine-to-machine authorization.
Gemini 3.1 Flash for real-time multimodal intake, function calling, and streaming review. Gemini 3.1 Pro for post-session economic audit. Two model tiers: Flash for speed and streaming; Pro for depth and executive narrative.
P402 router selects the right model tier per document or action. Budget caps enforced pre-execution. AP2 mandates govern per-session spending. ERC-8004 agent identity on Base mainnet. Full audit trail from extraction to onchain proof.
The three vertical demos stress different dimensions of the infrastructure: clinical governance, transactional routing, and multimodal volume. The enterprise demo proves the same infrastructure works as a horizontal cost management layer across every department, project, and employee simultaneously.
Per-token meter readings as the primary visual. Live ledger with running USDC cost. Economic audit panel as the climactic moment.
8-contract Acme/Beta data room. Flash for simple forms, Pro for complex agreements. Cross-document conflict detection. ABA-compliant audit trail on Tempo mainnet.
3 applicant scenarios: clean, income mismatch, likely fraud. Flash extracts 4 documents per applicant. Pro runs cross-document consistency. Fraud score with escalation threshold.
Org → dept → project → employee → token attribution. Routing optimization recommendations. Budget projections. Immutable session receipts. Synthetic dashboard live now.
The three vertical demos prove per-token billing works for specific regulated workflows. The enterprise demo proves it works as the metering layer underneath every workflow simultaneously. Together they prove P402 is complete AI cost infrastructure.
URAC (clinical), ABA (legal), HUD (housing), SOC2 (enterprise). Different audit artifacts, different compliance requirements. Same P402 session model and budget cap architecture satisfies all four.
Healthcare: $0.0000006/token. Legal: routing by document complexity. Real estate: per-applicant at scale. Enterprise: per-employee, per-project, per-department. One pricing model covers all four.
A health plan with P402 Enterprise uses the healthcare demo infrastructure inside the enterprise cost layer. Legal teams use the contract review infrastructure inside the same cost layer. The verticals are instances of the platform.
Clinical decisions for URAC. Contract reviews for ABA. Screening decisions for HUD. Employee AI usage for SOC2. The P402 ledger — with Tempo settlement hashes — provides the tamper-evident record all four require.
Every AI workflow — vertical or horizontal — has the same gap: no per-action cost attribution, no sub-cent settlement rail, no onchain audit trail. P402 fills it once. The four demos prove it from four different directions: clinical, transactional, volumetric, and organizational.
Real Gemini calls. Pre-funded wallet submits one real USDC.e TIP-20 transfer at session close. Tempo Explorer link live immediately.
Pre-recorded session stream. Real proof refs from a prior live run. All ledger events, approval gate, and economic audit run identically. No API keys needed to evaluate.