Use cases — 34 concrete patterns.
If you build agent tools, AI SaaS, or MCP servers, one of these is almost certainly your problem. Each page is a one-pager: the friction, the Spawnpay angle, the code, the path to first paid call.
- Bill agents per OpenAI call — USDC, no card, no chargebacks
Cards do not work for agents. Stripe needs a human, OpenAI invoices land at month-end, and 2.9% + 30¢ kills the unit eco… - Cap Claude API spend per agent — automatic USDC paywall
You hand an agent your Anthropic API key. It loops, makes 200,000 calls overnight, you wake up to a $4,000 bill. There i… - Charge Claude Code users for your MCP server — two lines
You built an MCP server. It works. People install it. You make zero. Stripe wants 7 pages of compliance and a card form … - Charge per LLM call without a Stripe integration
Per-call pricing on top of an LLM is the simplest pricing model for an agent tool. Implementing it via Stripe metered bi… - Payments for an MCP server — the missing piece of the spec
The MCP spec defines tools, prompts, and resources — but no payment primitive. Every MCP author re-invents billing badly… - Hard spending limits on AI agents — pre-charge enforcement
Agents loop. Loops cost money. A safety budget that is checked AFTER the call has already gone through is not a safety b… - BYOK proxy with built-in billing — one key, every provider
You want a unified billing pipeline across OpenAI, Anthropic, Brave, Replicate, ElevenLabs, etc. — but you do NOT want t… - Multi-provider LLM routing with one unified bill
Production agents use 3-5 LLMs depending on task. Tracking spend across OpenAI + Anthropic + Groq + Mistral + Together m… - OpenRouter alternative with per-agent budgets
OpenRouter is great for cross-model routing but has no native per-agent budget cap. Reselling tokens means you take on T… - LiteLLM alternative with on-chain receipts
LiteLLM is a routing proxy you self-host. Great unification, but you still need a billing layer on top — and you still n… - Give every AI agent its own USDC wallet on Base
Agents need money. Cards do not work. Bank accounts do not work. You end up sharing your personal card with a process th… - USDC micropayments for AI tools — fractions of a cent
Stripe minimum is $0.50. Coinbase Commerce minimum is $0.10. Card processors take 2.9% + 30¢ per call. None of this work… - Paywall without Stripe — works for agents, no card form
Stripe requires a card. Agents do not have cards. Card forms break automation. PCI compliance is overhead. Chargebacks l… - Agent-to-agent payments — send USDC by referral code
Two agents want to transact. Exchanging hex addresses is fragile (typos, copy-paste errors, no human-friendly form). Exi… - Sell access to your MCP server — distribute via npm
You spent a weekend building a useful MCP server. Maybe a few hundred installs. Zero revenue, because there is no way to… - Bill per Whisper transcription — AssemblyAI through Spawnpay
You have an agent that transcribes audio. AssemblyAI charges per audio-minute. You need to pass that cost (plus a margin… - Bill per ElevenLabs TTS call — paid-tts-mcp drop-in
TTS is expensive. Letting end-users invoke it without per-call accounting is a fast path to a runaway bill.… - Bill per DALL-E / Flux image generation
Image generation is expensive and easy to abuse. You need a billing primitive that enforces per-call cost in real time.… - Per-call billing for Brave Search — agent-friendly search
Brave Search has subscription pricing. You want pay-as-you-go for an agent that searches sporadically.… - Free Wikipedia + HN search through Spawnpay edge
Free APIs like Wikipedia's MediaWiki and HN Algolia work, but managing rate limits + retries across providers is repetit… - Bill per Groq inference — fast LLM through Spawnpay
Groq is fast and cheap, but multi-tenant billing on top of it requires a paywall layer.… - Mistral API with per-agent budget cap
Mistral usage scales fast in agent loops. You need a hard ceiling per agent per day.… - Bill per Cohere embedding — production-grade rate gate
Embeddings run in batches of thousands. A single misconfigured pipeline can spike usage 100x in minutes.… - Voyage AI embeddings — paid-embed-mcp drop-in
Voyage embedding pricing requires its own dashboard, its own invoicing flow.… - Perplexity API with USDC billing — agent-ready search
Perplexity's pay-per-call pricing is great, but you still need a way to gate end-user spend on your tool.… - DeepSeek API budget enforcement
DeepSeek is cheap, so agents tend to call it in loops. Cheap × big loop = expensive.… - Fireworks AI with on-chain receipts
Open-model inference at scale needs an audit trail finance can reconcile.… - Bill per Grok inference call (xAI)
Grok API is new. There is no native per-tenant billing primitive.… - Together AI paywall for multi-tenant tools
Open-model hosting at Together AI is cheap, but you need to bill end-users without renting Stripe.… - Stability AI image generation billing
Stability AI has per-image costs that vary by model. Tracking spend per end-user is a pain.… - Per-agent rate limit with USDC enforcement
Rate limiting by IP or key is brittle when agents share infra. You want economic rate limiting — the agent runs out of m… - Payments rails for an AI tool marketplace
You are building an AI tool marketplace. Stripe Connect onboarding is multi-week. KYC for sellers is multi-week. Card fe… - Charge for AI tools without taking a credit card
Card collection is a UX killer for AI tool consumption. Users abandon at the card form. PCI compliance is overhead. Char… - Unified LLM spend receipts across all providers
Finance wants one CSV with all LLM spend across OpenAI + Anthropic + Groq + Mistral + Together. Five dashboards, five ex…