Reference

Plans and limits

Seats, concurrency, channels and allowances by tier.

About 4 minutes

Five tiers, each priced two ways — lower with your own provider keys, higher with AI included.

TierSeatsConcurrencyChannels
Access111
Solo333
Team8810
Business202025
Scale40unlimited30

Current prices and AI credit allowances are on the pricing page, which is the authoritative source — figures here would drift.

What each limit actually constrains

Seats — how many people can use the workspace.

Concurrency — how much work runs at once. This is the number most people under-buy: it decides whether five objectives progress in parallel or queue behind each other. If work feels slow and nothing is erroring, this is usually why.

Channels — how many inbound connections you can have.

AI credit — the monthly managed-AI allowance. Irrelevant on bring-your-own-key.

Add-ons

Extra licences, channels, managed searches and external actions per day are available à la carte, so a single constraint does not force a tier change.

Gotchas

  • Concurrency is not speed. It is parallelism. One objective does not run faster on a higher tier; five run at once instead of queueing.

Checked against

  • helpers/pricingCatalog.ts
  • endpoints/billing/tiers_GET.ts