How we rate communication APIs
Methodology 2026.09-pre, published 23 September 2026. Status: preliminary — the weights will be calibrated once enough cards are verified, and every change gets a new version. The tables below are generated from the same file the scoring code reads.
What we check, and what we don't
Every fact comes from a public page — the vendor's pricing page, API reference, webhook documentation, trust center, official repositories — or from review platforms. We don't run hands-on tests or send messages through the providers, so no page on this site reports delivery results. Each fact records where it came from and when it was checked, and each card is re-checked at least every six months.
The evidence ladder
| Status | Meaning | Credit |
|---|---|---|
| Official docs | Vendor documentation, pricing page or official repository, linked to the section | full |
| Official, other | Vendor blog, changelog, marketplace listing or official MCP registry | full |
| Third-party confirmed | At least two independent sources where no official one exists | full |
| Vendor claim | Marketing claim without documentation | 50 % of the points |
| Not verified | Not found — treated as unknown | unknown |
Unknown data and coverage
When a fact is not verified, its criterion is left out and the remaining weights are scaled up, so a missing fact never counts as zero. Every provider gets a score; coverage, the share of the methodology weight backed by verified facts, is shown next to it, so a score built on few facts is visible as such. A country ranking lists only providers whose channel is available in that country. A ranking is published only when at least five providers are scored.
Overall rating
| Component | Weight | Criteria inside the component (share) |
|---|---|---|
| Pricing & Free Tier (average of the provider's channels) | 25 | Free access 25 · Scenario price 60 · Pricing transparency 15 |
| Developer & AI Readiness | 28 | MCP 27.5 · Agent skills 7.5 · Assistant apps 2.5 · Llms.txt 2.5 · OpenAPI 2.5 · AI docs 2.5 · Native AI features 5 · Scoped keys 5 · OAuth 16 · Webhook signing 7 · SDKs 8 · Test mode 8 · Idempotency 6 |
| Reputation & Reviews | 25 | Review platforms 100 |
| Channel Coverage | 10 | Channel groups 100 |
| Category Functionality (average of the provider's channels) | 8 | Channel-specific, see below |
| Localization & Regional Support | 2 | Languages 100 |
| Built-in AI Agent & Chatbot Deployment | 2 | Ready-to-deploy AI agent 100 |
Category rating
| Component | Weight | Criteria inside the component (share) |
|---|---|---|
| Pricing & Free Tier | 30 | Free access 25 · Scenario price 60 · Pricing transparency 15 |
| Developer & AI Readiness | 30 | MCP 27.5 · Agent skills 7.5 · Assistant apps 2.5 · Llms.txt 2.5 · OpenAPI 2.5 · AI docs 2.5 · Native AI features 5 · Scoped keys 5 · OAuth 16 · Webhook signing 7 · SDKs 8 · Test mode 8 · Idempotency 6 |
| Reputation & Reviews | 25 | Review platforms 100 |
| Category Functionality | 15 | Channel-specific, see below |
Country + channel rating
| Component | Weight | Criteria inside the component (share) |
|---|---|---|
| Pricing & Free Tier (local data first) | 25 | Free access 25 · Scenario price 60 · Pricing transparency 15 |
| Developer & AI Readiness | 25 | MCP 27.5 · Agent skills 7.5 · Assistant apps 2.5 · Llms.txt 2.5 · OpenAPI 2.5 · AI docs 2.5 · Native AI features 5 · Scoped keys 5 · OAuth 16 · Webhook signing 7 · SDKs 8 · Test mode 8 · Idempotency 6 |
| Reputation & Reviews (local data first) | 20 | Review platforms 100 |
| Local Market Fit | 15 | Channel availability 26.67 · Local payments 20 · UI docs language 13.33 · Support language 13.33 · Local phone 13.33 · Billing presence 13.34 |
| Category Functionality | 10 | Channel-specific, see below |
| Compliance Transparency | 5 | DPA 30 · Subprocessors 20 · Data residency 20 · AI data usage 10 · Local data-law page 20 |
Channel functionality and price scenarios
Transactional email
Price scenarios: 10,000 emails / month; 100,000 emails / month.
Free forever counts in full from 3,000 emails per month.
Functionality criteria: SMTP relay 12 · Inbound parsing 10 · Templates 10 · Event webhooks 14 · Suppression mgmt 10 · Dedicated IP 8 · Log retention 8 · Scheduling 6 · Attachments 4 · Payment notifications 8 · Native payment-provider integration 10
SMS
Price scenarios: 1,000 outbound SMS / month. GSM-7 single segment, outbound, domestic, the provider's cheapest compliant sender type for application-to-person traffic (US: 10DLC long code incl. number rental and campaign fees). Carrier pass-through fees are shown separately and excluded from the comparison because they are set by carriers, not by providers.
Free forever counts in full from 50 sms segments per month.
Functionality criteria: Two-way messaging 10 · Delivery receipts 14 · Sender types 14 · Number provisioning via API 8 · Sender registration support 10 · Opt-out handling 10 · MMS / RCS 6 · Scheduling 5 · Segment and Unicode rules documented 5 · Payment links in messages 8 · Native payment-provider integration 10
OTP & verification
Price scenarios: 1,000 SMS verifications / month. SMS channel, successful verifications, domestic
Free forever counts in full from 50 verifications per month.
Functionality criteria: Channels sms 10 · Channels whatsapp 10 · Channels voice 6 · Channels email 6 · Managed codes 14 · Fraud protection 18 · Rate limiting 10 · Silent network auth 8 · Localized templates 8 · Analytics 10
WhatsApp Business API
Price scenarios: 1,000 utility template messages / month; 10,000 utility template messages / month. Utility-category template messages to recipients in the market, business-initiated. Meta's own per-message fee is identical across providers and shown separately; the comparison uses the provider's platform fees (subscription + per-message markup).
Free forever counts in full from 1,000 whatsapp messages per month.
Functionality criteria: Official Meta partner 10 · Embedded signup 8 · Template API 14 · Interactive messages 10 · Media messages 5 · Flows 8 · Inbound webhooks 10 · Shared inbox 5 · Payment links in messages 10 · In-chat payments 10 · Native payment-provider integration 10
Prices
A provider's price on a scenario is compared with the median of all priced providers for the same channel and country: at the median it earns 60 % of the price points, at half the median the full points, at double the median 20 %. A comparison needs at least five priced providers; otherwise the criterion is unknown. Carrier and Meta fees that every provider passes through are shown but left out of the comparison. Displayed prices stay in the vendor's currency.
Reviews
We record the rating, scale, number of reviews and date from each review platform, and whether the listing covers the company or a specific product. Trustpilot carries half of the combined weight and G2 and Capterra a quarter each; platforms with fewer than 100 reviews count proportionally less. Small review bases are pulled toward a neutral midpoint, so the number of reviews gives a bounded bonus. Below ten reviews in total the criterion is unknown. Review texts are never copied.
Payments inside conversations
Payments are not a separate rating. Payment links in messages, in-chat payments, payment notifications and native payment-provider integrations are criteria inside each channel's functionality. We don't compare payment providers' fees and the site never processes payments.
Neutrality
Scores are computed from data and the published weights only. No provider pays for placement or sees its score before publication. See the editorial policy and the corrections log.
Changes to this methodology
Every change to weights or rules is listed here with its date, and all scores are recomputed from the same facts when it lands.
- 24 Sep 2026: Any OAuth 2.0 grant documented for the API counts in full, client credentials included; OAuth that exists only for a provider's MCP server counts as partial. An llms.txt that an SEO plugin generates for a marketing site and that links no API documentation does not count. A general cloud MCP server that can call the provider's API earns the same points as a product-specific one.
- 24 Sep 2026: Webhook replay protection means an ID, timestamp or nonce inside the signed payload and documentation that tells you to reject old or repeated deliveries.
- 24 Sep 2026: Developer & AI readiness: MCP server 22.5 → 27.5 and official agent skills 5 → 7.5 of the component; listings in the Claude and ChatGPT app directories 10 → 2.5, because a server or skill can be installed without a directory listing.
- 24 Sep 2026: Transactional email pricing: the free tier weighs 50 of the pricing component (was 25) and is graded by monthly volume, from 0.4 at 100 emails a month to 1.0 at 10,000 on a log scale; a trial or credits earn 0.2.
- 24 Sep 2026: OTP & verification covers managed verification APIs and providers that document sending one-time codes over SMS or WhatsApp; providers without a verification API score 0 on managed codes, fraud protection and rate limiting.
- 24 Sep 2026: Developer & AI readiness: OAuth 2.0 6 → 16, scoped API keys 10 → 5, webhook signing 12 → 7. Every provider is scored; coverage is shown next to the score (the 80 % coverage minimum was dropped).