Frequently Asked Questions

SettleGrid is the settlement layer for the AI economy. It lets developers monetize any AI service — MCP tools, REST APIs, AI agents, model endpoints — with one SDK and one unified billing, metering, and payout system. Think of it as "Stripe for AI services" with real-time metering, multi-protocol support, and automatic revenue splits.

Sign up for a free developer account, connect your Stripe account for payouts, create a tool with a unique slug and pricing configuration, then install the @settlegrid/mcp SDK and wrap your handler. You can be live in under 5 minutes.

Yes. The Free plan is $0 forever with no catch — unlimited tools, 50,000 operations per month, per-call billing, a full dashboard, and a progressive take rate starting at 0% on your first $1K/mo of revenue. No credit card required. Most developers will never need to upgrade.

SettleGrid is protocol-agnostic. The @settlegrid/mcp SDK adds native billing to MCP tool servers and a REST middleware for any HTTP service. The hosted Smart Proxy brokers payments across 9 protocols: MCP (Model Context Protocol), x402 (Coinbase / Linux Foundation), Stripe MPP (Machine Payments Protocol — Stripe + Tempo, pending GA), AP2 (Google Agent Payments), ACP (Agentic Commerce Protocol — OpenAI + Stripe), UCP (Universal Commerce Protocol — Google + Shopify), Visa TAP (Token Agent Payments), Mastercard Verifiable Intent, and Circle Nanopayments (USDC). Detection adapters are in place for L402 (Bitcoin Lightning) and KYAPay (Skyfire + Visa). Additional rails are tracked as upstream specs mature: ACTP (Alipay's Agentic Commerce Trust Protocol, Ant Group), EMVCo agent payments, and DRAIN (Bittensor Subnet 58). One SDK covers the brokered set.

Any tool with 'active' status automatically appears in the Showcase. Set your tool to active from the Tools page in your dashboard.

Fork any of 1,017 open-source MCP servers on GitHub, add SettleGrid billing, deploy to Vercel or Railway. Your tool automatically appears in the Showcase and becomes discoverable via the Discovery API and MCP Discovery Server.

Run npx create-settlegrid-tool in your terminal. It generates a complete MCP server project with SettleGrid billing, tests, Dockerfile, and README. Choose from 4 templates (blank, rest-api, openapi, mcp-server) and 3 deploy targets (Vercel, Docker, Railway). The generated project includes pricing configuration, error handling, and CI already wired in — you can be live in minutes.

SettleGrid provides two discovery mechanisms. First, the MCP Discovery Server: run npx @settlegrid/discovery to start a local MCP server that any AI client (e.g., Claude Desktop) can connect to. It exposes search_tools, get_tool, list_categories, and get_developer tools so agents can find and evaluate SettleGrid-powered services at runtime. Second, the Discovery API: public REST endpoints at /api/v1/discover let any HTTP client search tools by keyword, category, or developer. Both require no authentication.

The SDK uses an in-memory LRU cache for key validation (5-minute TTL) and fires metering requests asynchronously via Redis DECRBY on the hot path. Typical overhead is under 10ms — well below the threshold users would notice.

Yes. The pricing config supports per-method overrides. Set a defaultCostCents for all methods, then override specific ones in the methods map. For example, a simple lookup might cost 1 cent while a complex analysis costs 10 cents.

Yes. While the package is called @settlegrid/mcp, it includes a settlegridMiddleware() function for REST APIs (Express, Fastify, etc.). The SDK's wrap() function works with any async handler regardless of protocol.

No. While MCP is our primary focus, the SettleGrid SDK wraps any function handler. It works with REST APIs, Express routes, Next.js API routes, serverless functions, and any callable endpoint. If your code accepts input and returns output, SettleGrid can meter and bill it.

Six models: per-call (a fixed cost per call — named per-invocation in the dashboard and API), per-token, per-byte, per-second, tiered (different prices per method), and outcome-based (charge only on success).

Your tools keep working beyond 50,000 ops/month — we never cut off consumers. The progressive take rate applies to overage operations. Upgrade to Builder ($19/mo) or Scale ($79/mo) for higher limits.

It depends on your tool's value and compute costs. Here are benchmarks by tool type: - Simple lookups/search: 1-5 cents per call - Data enrichment/APIs: 5-25 cents per call - AI-powered analysis: 10-50 cents per call - Complex multi-step workflows: 25 cents - $1+ per call Start on the lower end to attract early users, then adjust based on demand and feedback.

Consumers purchase credits via Stripe (credit card). When they call your tool, credits are deducted in real-time. You receive payouts via Stripe Connect on your chosen schedule (daily, weekly, or monthly).

No. Credits purchased for a specific tool never expire and can be used at any time.

Invocations return an InsufficientCreditsError (HTTP 402) with the required and available amounts until you purchase more credits. Your tool continues to work for other consumers who have credits.

Revenue from your tools accumulates in your SettleGrid developer balance. Payouts are disbursed via Stripe Connect Express to your linked bank account. You can choose weekly or monthly payout schedules, or trigger a manual payout from the dashboard at any time.

All plans use a progressive take rate based on monthly tool revenue: 0% on the first $1,000/mo (you keep 100%), 2% on $1,001-$10,000, 2.5% on $10,001-$50,000, and 5% above $50,000. Most developers pay 0%. Need a custom arrangement? Email support@settlegrid.ai.

Yes. Consumers who have used a tool at least once can leave a 1-5 star rating with an optional written review (up to 1,000 characters). One review per consumer per tool.

Yes. Developers can view and respond to all reviews from the Reviews tab in the dashboard. Responses appear publicly on the tool's storefront.

Before a tool can be activated and appear in the Showcase, it must meet minimum quality requirements: description of at least 50 characters, pricing configured, category selected, and developer profile complete.

Tools earn the Verified badge automatically after receiving their first real (non-test) invocation. It signals to consumers that the tool has been successfully called in production.

Go to the Consumer Dashboard, find the tool you purchased credits for, and click 'Generate API Key'. The full key is shown once — copy it immediately.

Yes. The Budget Controls section of the Consumer Dashboard lets you set daily, weekly, or monthly spending limits per tool with alert thresholds.

API keys are SHA-256 hashed before storage — SettleGrid never stores plaintext keys. Only the first few characters (the key prefix) are stored in clear text so you can identify keys in the dashboard. The full key is shown only once at creation time.

SettleGrid runs a three-signal fraud detection system: (1) rate spike detection flags abnormal invocation bursts, (2) new-key velocity checks flag high-value usage from newly created keys, and (3) rapid duplicate deduplication catches repeated identical requests. Suspicious invocations are flagged and can trigger email alerts.

Every webhook delivery is signed with HMAC-SHA256 using a per-endpoint secret. The signature is included in the X-SettleGrid-Signature header. Verify it by computing the HMAC of the raw request body using your secret and comparing it to the header value.

We automatically index tools from MCP registries, npm, Hugging Face, and Replicate. If your tool appears as 'unclaimed', you can claim it and set pricing in 2 minutes. Check your email for a claim link, or search for your tool on the Explore page at settlegrid.ai/explore.

SettleGrid crawls MCP registries, npm, and other sources every 6 hours. When we find a tool that uses @settlegrid/mcp or matches an MCP server pattern, we create an unclaimed listing and email the developer (resolved from GitHub commits or npm profiles). Click the claim link in the email, verify ownership, and start earning.

SettleGrid is a universal billing layer for any AI service. Supported service types include: LLM inference proxies (OpenAI, Anthropic, Cohere, Mistral), browser automation tools (Playwright, Browserbase, Puppeteer), media generation services (DALL-E, Stable Diffusion, ElevenLabs), code execution sandboxes (E2B, Modal, Fly Machines), data enrichment APIs, MCP tools, agent-to-agent orchestration workflows, communication services (Twilio, Resend, SendGrid), search and retrieval (RAG pipelines, vector search), and compute infrastructure (GPU instances, model hosting). If your service accepts input and returns output, SettleGrid can meter and bill it.

Yes. SettleGrid provides comprehensive agent budget controls through three features: (1) Transaction Explorer gives full visibility into every AI service call with cost breakdowns, (2) Smart Proxy enforces budget limits in real-time and returns HTTP 402 when budgets are exceeded, and (3) progressive budgets let you set daily, weekly, or monthly spending limits per agent, per tool, or per workflow. Combined with Agent Identity (KYA) verification, you can delegate budgets to autonomous agents with full control and audit trails.

Stripe is a payment processor. Orb is a usage-based billing platform for SaaS. SettleGrid is an AI-native settlement layer purpose-built for the AI economy. Key differences: (1) Discovery — SettleGrid includes a built-in marketplace, Discovery API, and MCP Discovery Server so AI agents can find and pay for your services automatically. (2) Metering — real-time metering with atomic Redis balance checks on the hot path, not batch billing. (3) Multi-protocol — the Smart Proxy brokers 9 agent payment protocols (MCP, x402, Stripe MPP, AP2, ACP, UCP, Visa TAP, Mastercard Verifiable Intent, Circle Nanopayments) with detection adapters for 2 more (L402, KYAPay) and tracks 3 emerging rails (ACTP, EMVCo agent payments, DRAIN), versus Stripe/Orb's REST-only approach. (4) Progressive pricing — 0% take rate on your first $1K/mo. (5) Agent-native — budget delegation, KYA identity, multi-hop settlement for autonomous agent workflows.