# Interrupt Index

> Independent, probe-based reviews of how agent-friendly tools, APIs and sites are.
> Core metric: the Interrupt Score — every point where a human must intervene for an
> autonomous agent to proceed. 68 listings scored.

This file is the agent manual. We publish it for completeness and grade it *emerging*
on the sites we review — prefer the JSON API below, the canonical machine surface.
The full guide for agents (with copyable install commands) is at https://interruptindex.com/agents
and its machine-readable twin is https://interruptindex.com/api/agent/install.

## Interfaces (all unauthenticated)

- [Listings JSON](http://127.0.0.1:3001/v1/listings.json): the full scored catalog — composite, badge, machine-readability, usage mode. Supports ?category=, ?usage_mode=, ?limit=.
- [Listing detail JSON](http://127.0.0.1:3001/v1/listings/fewsats.json): full findings + per-check breakdown for one tool (replace the slug).
- [Categories JSON](http://127.0.0.1:3001/v1/categories.json): the category taxonomy.
- [Stats JSON](http://127.0.0.1:3001/v1/stats.json): site-wide aggregate counters.
- [OpenAPI 3.1](http://127.0.0.1:3001/v1/openapi.json): the formal contract for every endpoint above.
- [MCP server manifest](https://interruptindex.com/.well-known/mcp.json): stdio lookup_tool — query a score at decision time. Install: npx interrupt-index-mcp@latest setup (package pending npm publish; runs from source until then).
- [Setup contract JSON](https://interruptindex.com/api/agent/install): every interface, the score model, and guardrails as data.
- [Submit a site](https://interruptindex.com/api/submit): POST {"url"} to queue scoring. Rate limit 5/hour/IP.

## Using this site as an agent

1. Discover: fetch /v1/listings.json ONCE per session and filter locally. Do not poll it
   per loop-iteration. Scores are comparable only within a usage_mode (integrated vs
   transactional) — never rank across the two.
2. Inspect: for each shortlisted tool, GET /v1/listings/{slug}.json for its findings.
   Check the findings before any irreversible action (signup, payment, subscription).
3. Verify: read the confidence label and lastProbedAt. Treat a Provisional or stale score
   as a triage signal, not a live guarantee. A Blocked badge or a tos_bot_ban finding
   means the vendor has signalled bots are not welcome — surface it, don't push through.

## Score model

- composite = round( (interruptScore + machineReadabilityScore) / 2 ) — a flat 50/50 mean.
- interruptScore (0–100): how few points force a human into the loop. This alone sets the badge.
- machineReadabilityScore (0–100): how legible the integration is. Reported alongside, never in the badge.
- badges: Zero-Interrupt / Low-Interrupt / Moderate-Interrupt / High-Interrupt / Blocked.
  A ToS bot-ban, contact-sales-only, or any hard per-transaction step forces Blocked regardless of score.
- confidence: Verified (Probed) / Verified (Partial) / Verified (Documented) / Provisional; decays with probe age.
- Full derivation: https://interruptindex.com/methodology

## Docs

- [For agents (install guide)](https://interruptindex.com/agents): every interface, copyable install commands, cost discipline.
- [Methodology](https://interruptindex.com/methodology): scoring algorithm, badge tiers, evidence grades.
- [Directory](https://interruptindex.com/directory): all scored tools.
- [Resources](https://interruptindex.com/resources): agent-usability standards and research.
- [Research](https://interruptindex.com/research): the evidence behind agent-readiness.

## Top listings

- [OpenWeatherMap](https://interruptindex.com/tools/openweathermap): LOW_INTERRUPT, overall 93/100
- [x402 (Coinbase protocol)](https://interruptindex.com/tools/x402-coinbase-protocol): LOW_INTERRUPT, overall 89/100
- [Fewsats](https://interruptindex.com/tools/fewsats): ZERO_INTERRUPT, overall 88/100
- [ChatGPT Instant Checkout / OpenAI ACP](https://interruptindex.com/tools/chatgpt-instant-checkout-openai-acp): MODERATE_INTERRUPT, overall 82/100
- [example.com](https://interruptindex.com/tools/example-com): LOW_INTERRUPT, overall 75/100
- [Hetzner](https://interruptindex.com/tools/hetzner): HIGH_INTERRUPT, overall 54/100
- [Ticketmaster](https://interruptindex.com/tools/ticketmaster): BLOCKED, overall 8/100
- [Fly.io](https://interruptindex.com/tools/fly-io): unscored
- [Vultr](https://interruptindex.com/tools/vultr): unscored
- [AWS](https://interruptindex.com/tools/aws): unscored
- [DigitalOcean](https://interruptindex.com/tools/digitalocean): unscored
- [Skyfire](https://interruptindex.com/tools/skyfire): unscored
- [Crossmint](https://interruptindex.com/tools/crossmint): unscored
- [Google Cloud Platform](https://interruptindex.com/tools/google-cloud-platform): unscored
- [Stripe Agent Toolkit MCP Server](https://interruptindex.com/tools/stripe-agent-toolkit-mcp-server): unscored
- [Payman AI](https://interruptindex.com/tools/payman-ai): unscored
- [Prixe](https://interruptindex.com/tools/prixe): unscored
- [x402 Bazaar](https://interruptindex.com/tools/x402-bazaar): unscored
- [Modal](https://interruptindex.com/tools/modal): unscored
- [Nevermined](https://interruptindex.com/tools/nevermined): unscored

---o---