Product

Tempo MPP (Machine Payments Protocol)

Stripe + Paradigm's agent-native payment protocol and settlement chain; session-based authorization + on-chain aggregation, aiming to be the "per-call settlement layer" for the AI agent era.

1. Core Product / Service

Tempo MPP is not Stripe in general but a dedicated payment chain + protocol incubated by Stripe in partnership with Paradigm [tempo.xyz, 2026-05-09]:

  • MPP (Machine Payments Protocol): open protocol reviving HTTP 402 "Payment Required", letting agents auto-negotiate payment when calling APIs; co-authored by Stripe and Tempo Labs, extensible across chains / payment methods.
  • Tempo chain: mainnet live 2026-03-18 (Chain ID 4217), ~500ms deterministic finality, sub-cent fees, optimized for high-frequency low-amount agent payments; not a general-purpose smart-contract L1.
  • Sessions model: "OAuth for money" — authorize once, charge many; thousands of small charges aggregated into a single settlement, avoiding the cost of putting every transaction on-chain.
  • Settlement assets: USDC (on-chain) + Visa Shared Payment Token (fiat side); the MPP protocol itself is asset-agnostic.
  • Service directory: *.mpp.tempo.xyz native endpoints (Stripe direct integrations) + third-party aggregators (locus etc.) extending to 100+ services.

Pricing from Jimmy's live tests (2026-04 [local]):

Endpoint Type Wallet charge
Perplexity /search Native search ~$0.007/call
Perplexity /chat (Sonar) LLM + citations ~$0.015/call
OpenRouter gpt-4o-mini LLM ~$0.00075 base + token
SerpAPI Search per-call small amount

2. Target Users & Pain Points

  • API providers: want to sell agents per-call instead of $X/month subscriptions, and don't want to integrate crypto / billing / KYC themselves. Plug in an MPP endpoint, receive USDC / Visa.
  • Agent developers: need on-demand pay-as-you-go inside an agent — web search, LLM, data enrichment, browser automation — instead of opening an account and binding a card with each supplier.
  • Pain point: traditional Stripe Connect has minimum fees of ~$0.30 + 2.9%, can't do sub-cent micropayments; credit cards also can't support high-frequency agent loops. MPP opens up this path with session aggregation + on-chain sub-cent fees.

3. Competitive Landscape

Protocol Controller Settlement Finality Model Notes
Tempo MPP Stripe + Paradigm USDC on Tempo / Visa SPT ~500ms Session authorization + aggregation Built-in compliance stack, enterprise-friendly
x402-protocol Coinbase + Cloudflare + Stripe (open) USDC on Base / Solana On-chain instant Single 402 challenge / no account Apache 2.0, no vendor lock-in
skyfire Skyfire (independent) USDC + fiat API gateway KYA agent identity + payments Takes the API-gateway route rather than protocol-first
kite-ai Kite (PayPal Ventures + GC) KITE on Avalanche subnet On-chain Agent Passport + AIR L1 with built-in identity layer; mainnet Q1 2026
Stripe Connect Stripe Fiat T+2 Platform/merchant High minimum fees, unsuitable for agent micropayments

Differentiation: Tempo MPP's wedge is "protocol + own settlement chain + Stripe compliance install base" three in one; x402 is lighter and more open but lacks compliance and fiat rails; Skyfire leans SaaS gateway; Kite bets on its own L1 and identity layer. WorkOS's judgment (2026-05-09): x402 and MPP are complementary and may coexist — the former for "discrete one-off calls", the latter for "high-frequency agent sessions".

4. Unique Observations

  • Early stage, TVL very small: live test on 2026-04-02 showed TVL only ~$31,102 [local]. The design partner list (Visa, Mastercard, OpenAI, Anthropic, Shopify, DoorDash, Ramp, Deutsche Bank, Standard Chartered, Revolut, Nubank) is impressive, but actual on-chain volume is still in seeding.
  • Locus is currently the largest 3p aggregator: locus (YC F25, paywithlocus.com) signs 50+ APIs in bulk, wraps them into MPP-compatible endpoints, lists them on Tempo, and adds a 2-3x markup (live test: Perplexity chat $0.005 → $0.015). Tempo native endpoints ~20, Locus endpoints ~50+. High short-term value but significant long-term risk of being squeezed by Tempo native integrations [local: jclaw-2026-04-04].
  • Perplexity's $0.007/search on Tempo is currently the cleanest agent web-search channel — Jimmy has switched OpenClaw's default web_search to the tempo-perplexity proxy (path: OpenClaw → 127.0.0.1:18800 → tempo CLI → Perplexity /search → on-chain USDC.e charge).
  • KV cache passthrough is an open question: whether the Tempo MPP OpenRouter endpoint (openrouter.mpp.tempo.xyz) can actually pass upstream KV cache discounts to the caller is unconfirmed in live testing; if not passed through, the cost impact on long-context agents is large.
  • Blue-ocean directions on Tempo (Jimmy's view [local: 2026-04-02]): (1) IM messaging services (biggest gap, OpenClaw already has the capability); (2) Chinese-data APIs (Chinese search, A-share, Qichacha-class — 100% gap on Tempo); (3) multi-tool agent task execution pay-per-call.
  • Next round of competition: Tempo needs to convert design partner names into endpoints that actually generate transaction volume; the standards battle between MPP and x402-protocol is not yet settled and may coexist — Stripe also supports x402 in stripe-agent-payment, betting on both sides.

5. Financials

  • Parent Stripe: private company, 2024 valuation $50B+; MPP is one product line of Stripe's agent payment strategy.
  • Tempo Labs (independent entity, incubated by Stripe + Paradigm): valuation ~$5B, financing ~$500M [local: 2026-04-02-diary-claudecode].
  • No native token publicly issued (2026-05-09 mainnet announcement did not mention).

6. People & Relationships

Last compiled: 2026-05-09