Published 2026-09-27 · Tested 2026-09-27

Metronome

A-

Metronome received 9 PASS votes and passed 5 of five agent surface checks. The clearest finding came from the find the exact limits task.

Panel: GPT 5.6 Sol, Opus 5, DeepSeek v4F Battery: v1 Read as markdown (opens in a new tab)

Three AI models, GPT 5.6 Sol, Claude Opus 5, and DeepSeek v4 Flash, each read Metronome’s public documentation independently and attempted five first-hour developer jobs: make the first API call, find the exact limits, recover from a 429, verify a webhook, use the Python SDK.

No accounts, API calls, or code execution were used. Every verdict came from public pages and every published quotation passed a live verification check.

Freshness

How rechecks work
Tested
Quotes verified
Surface rechecked
Not yet rechecked
Metronome Mintlify · published
A-

92.5% · 74/80 · AI Agent Readiness Score · reading 30 pts · surface 50 pts

llms.txt PASS
llms-full.txt PASS
markdown mirror PASS
MCP server PASS
docs AI PASS
Task GPT 5.6 SolOpus 5DeepSeek v4F Consensus
Make the first API call PARTIALPASSPASS PASS
Find the exact limits PARTIALPARTIALPASS PARTIAL
Recover from a 429 PASSPASSPASS PASS
Verify a webhook PARTIALPASSPASS PASS
Use the Python SDK PARTIALPARTIALPASS PARTIAL

docs platform: Mintlify (unscored) · verified 2026-09-27

What to fix first

These 2 fixes could add up to 4 points to the AI Agent Readiness Score. The list ranks each fix by the points it would add. How the ranking works

  1. 1
    +2 points Find the exact limits PARTIAL

    Found: The Limits page gives a 2,750 requests-per-second cap, while ingest pages cite 100,000 and 110,000 events per second.

    Fix: State ingest throughput once on the Limits page, in requests and events per second, and link other pages there.

    Evidence: docs.metronome.com/api-reference/rate-limits (opens in a new tab)

  2. 2
    +2 points Use the Python SDK PARTIAL

    Found: The Python SDK ingest example sends tokens as an integer, while the usage events guide requires string property values.

    Fix: Quote every properties value as a string in the Python SDK ingest example.

    Evidence: docs.metronome.com/api-reference/sdks (opens in a new tab)

What the docs get right

  • Recover from a 429: 3 PASS votes. The status and ingestion pages identify client versus customer scope, then prescribe delayed retries with exponential backoff.
  • Make the first API call: 2 PASS votes. The guides API quickstart carries the token prerequisite, the sandbox signup link, a runnable curl against https://api.metronome.com/v1/ingest with the Bearer header, and a Step 7 block that explains what a 200 means and how to verify the event in the dashboard.
  • 5 of 5 agent surface checks. Present: llms.txt, llms-full.txt, markdown mirrors, an MCP server, docs AI.

Make the first API call

PASS

PASS consensus from 2 PASS, 1 PARTIAL.

The quickstart joins token creation to /v1/ingest, but its March 2026 timestamps exceed the documented 34-day window on the test date. The guides API quickstart carries the token prerequisite, the sandbox signup link, a runnable curl against https://api.metronome.com/v1/ingest with the Bearer header, and a Step 7 block that explains what a 200 means and how to verify the event in the dashboard. The quickstart links to the send-usage-events guide which documents the /ingest endpoint structure, and the SDK README provides the ingest call code.

Find the exact limits

PARTIAL

PARTIAL consensus from 1 PASS, 2 PARTIAL.

The limits page gives the 2,750 global cap and per-environment platform defaults, while two ingest pages claim 100,000 and 110,000 events per second. The Limits page states the global cap of 2,750 req/sec, an ingest tier of 1,100 req/sec, and ten per-object platform limits in one table, but three other pages give ingest ceilings and defaults that do not reconcile with it: 110,000 events/sec with a 5,000 events/sec default on the scale guide, and 100,000 events/sec on the ingest reference. Values are stated in one authoritative page with tiered API rate limits and a separate platform-limits table. No contradictions found across pages.

Recover from a 429

PASS

PASS consensus from 3 PASS.

The status and ingestion pages identify client versus customer scope, then prescribe delayed retries with exponential backoff. The status codes page names the cause, documents the X-Metronome-Rate-Limit-Type header with its client and customer values, and the send usage events guide adds exponential backoff plus the assurance that transaction_id makes retries safe. The docs name the cause, recommend backoff with exponential retry, and the Python SDK auto-retries 429 errors by default.

Verify a webhook

PASS

PASS consensus from 2 PASS, 1 PARTIAL.

The page documents HTTPS receipt and HMAC-SHA256 verification, but only its Slack subsection states the destination-creation steps. The page covers destination setup, payload shapes per webhook type, source IPs, deduplication, the five-minute staleness rule, the bytes-not-parsed-JSON warning, and a worked example whose published signature I recomputed to a byte-exact match; the bash snippet is the one weak spot, since echo -n does not expand \n without -e. End-to-end process documented: webhook destination setup, event receipt, acknowledgment, signature verification with HMAC-SHA256, and code examples in bash, JavaScript, and Go.

Use the Python SDK

PARTIAL

PARTIAL consensus from 1 PASS, 2 PARTIAL.

The SDK install and ingest example are clear, but its numeric tokens property conflicts with the usage guide’s string-only rule. The SDK page gives pip install --pre metronome-sdk, the from metronome import Metronome constructor, and a client.v1.usage.ingest example, but that example sends "tokens": 1000000 as an integer while the usage events guide instructs that all property values be strings, so the canonical example contradicts the canonical rule. Minimal working example matches the API form shown on the docs (POST /v1/ingest with the same event schema).

The receipt

All limits are measured in requests per second. Exceeding a limit returns a 429 Too Many Requests response.

The limits page gives the 2,750 global cap and per-environment platform defaults, while two ingest pages claim 100,000 and 110,000 events per second.

Agent surface notes

JSON-RPC initialize POST returned protocolVersion 2025-06-18 and serverInfo name Metronome version 1.0.0 with search/retrieval tools for the site.

After hydration the docs header shows an Ask Assistant control; one click opened the Mintlify assistant panel with an Ask a question... input.

Show the score

AI Agent Readiness Score 92.5%, grade A-

Paste this into a readme:

[![AI Agent Readiness Score 92.5%](https://docsforagents.com/badge/metronome.svg)](https://docsforagents.com/reports/metronome-docs-ai-agent-readiness/)

Method note

This is a reading test of public documentation, not an execution test. No accounts were created and no API calls were run. The AI Agent Readiness Score counts fifteen reading votes at PASS 2, PARTIAL 1, and FAIL 0, for 30 possible points. Five agent surface checks add 10 points each. The total is 80. Consensus chips show each row majority and do not affect scoring. The panel split on 4 of five tasks. Quotes shown here were re-fetched and confirmed verbatim on 2026-09-27.

Read the full methodology

Put another docs site through the battery.

Nominate a docs site