mepa8
Docs · Connect

Connect ElevenLabs.

One scoped key, two read endpoints, a guided kill. Five minutes, most of it in the ElevenLabs dashboard.

What Mepa8 reads

The ElevenLabs connector is a reader of two endpoints and nothing else. The subscription endpoint returns the credits used and the limit, the reset time, the plan tier, the status and the current overage. The usage endpoint, GET /v1/usage/character-stats, takes a start and end time in milliseconds and a breakdown_type, and Mepa8 asks for product_type so a batch text-to-speech job is named as TTS on the day it runs. That page notes the endpoint is deprecated in favour of a workspace analytics query; Mepa8 will move when the replacement reports the same breakdown.

Both calls carry the key in the xi-api-key header, which is the only header ElevenLabs accepts for API keys per its authentication reference. Mepa8 makes 2 calls per check, 24 checks a day on the hourly schedule, so 48 requests a day against a key that cannot generate a single character of speech.

Create the key

  1. In the ElevenLabs dashboard, open the API keys page for your workspace and create a new key. Name it mepa8-read so it is obvious in the usage breakdown by key later.
  2. Restrict it. The authentication reference describes three restrictions a key can carry: endpoint scope, a credit quota, and an IP allowlist. Limit the scope to the user and usage endpoints, and set the credit quota to 0 so the key cannot spend even if it leaks.
  3. Copy the key once. ElevenLabs shows it at creation and Mepa8 will never show it back.
  4. In Mepa8, Add vendor → ElevenLabs → paste → Save. The first check runs immediately and the row shows used credits against the limit.
If scope restriction is not offered on your plan

Create the key anyway, set the credit quota to 0, and note on the vendor card that the key is unscoped. The quota is the control that matters for a reader.

What the row shows

On our own account the first check on October 8, 2026 returned 4,091,604 credits used of a 3,831,927 limit, status active, overage $46.55, reset on October 12. The usage breakdown by product type for the period showed 2,968,884 credits of text-to-speech, 803,780 of conversational AI and 365,059 of the LLM behind it, which is how one re-rendered audio course was named as the cause within a minute of connecting. Those figures are what a single scoped key reads.

“Your API key is a secret.”ElevenLabs authentication reference, read 2026-10-09

Alerts this connector raises

  • Warning at 80% of the credit limit, with the reset date.
  • Critical past 100%, stating whether usage-based billing is on, because with it off the API stops answering and a phone agent refuses every call.
  • Critical on any subscription status other than active, which is the past-due case that took our phone lines down on September 13, 2026.
  • Warning when one product type uses more than half the period's credits in a single day, which is the batch-job signal.

The kill, and why it is guided

ElevenLabs publishes no API call that revokes a key, lowers a quota or pauses a plan; the authentication reference describes restrictions you set when creating a key, not a way to disable one remotely. So the Mepa8 kill for ElevenLabs is the two-step flow with a person in it. Propose shows the blast radius you wrote on the card, for us every phone number at once, and the dashboard steps. Confirm with the single-use token records the decision and walks you through revoking the spending key or dropping its credit quota in the ElevenLabs dashboard. Restore is creating a key again and saving it. The read key Mepa8 holds is never the one you revoke, so monitoring keeps running through the kill.

Questions people ask

Straight answers

Does Mepa8 need a key that can generate speech?

No. Set the credit quota on the Mepa8 key to 0 and scope it to the user and usage endpoints. The connector makes 2 read calls per check and 48 a day; none of them can create audio.

Why is ElevenLabs Watched tier and not Full?

Because the kill cannot be an API call. ElevenLabs documents no endpoint to revoke a key or pause a plan, so the kill is guided: propose, confirm with a token, then the dashboard step. Cost and health are by API every hour, which is the Watched definition.

What if usage-based billing is on?

Then the API keeps answering past the limit and overage lands on the next invoice, which on our account was $46.55 for 259,677 credits over in September 2026. Mepa8 states in the critical alert whether overage billing is on so you know which failure you are facing: a bill or an outage.

Sources

  1. ElevenLabs — authentication (API keys, restrictions) — read 2026-10-09
  2. ElevenLabs — get user subscription — read 2026-10-09
  3. ElevenLabs — usage character stats — read 2026-10-09

Change log: October 9, 2026 — First published. Endpoint facts read from the ElevenLabs API reference the same day.

Start here

Find out what you are paying before the next invoice does.

Five business days. Every vendor listed with real cost, every leak named with its fix, and a board you keep.

Book a vendor audit →See the plansNo card on the form. A person replies within one business day.