Usage and limits
Your plan covers your book of accounts. Agents work every account in it, as often as the signals change, and you pay one flat monthly price for that coverage — there is no per-seat fee and no per-run charge to budget for.
Almost everything Sonora does costs nothing to run again and again: querying your own meetings, product usage, support history, and notes is free, no matter how often agents, chat, or actions read them. The only work Sonora meters is the small slice that has a real outside cost — reaching third-party data providers and the public web.
What’s free and what’s metered
Section titled “What’s free and what’s metered”| Work | Metered? |
|---|---|
| Querying your own data — meetings, product usage, issues, emails, notes, engagement | Free |
| A web search step inside an agent run | Metered |
| Looking up a person or company from an external data provider | Metered |
A health agent scoring an account from meetings and usage is free. A contact-enrichment agent that finds a LinkedIn profile and current title reaches an outside provider, so it counts against your plan’s external-work allowance.
You don’t manage this per run. You write the agent, choose its data sources, and Sonora handles the accounting.
Reading your usage
Section titled “Reading your usage”Settings → Usage shows external work used this period against your plan’s allowance, and the date the count resets. It resets every billing period on your contract anniversary day: if your anniversary is the 15th, each period runs from the 15th of one month to the 15th of the next.
Nothing switches off mid-run. As you approach or pass your allowance, your account team reaches out to right-size the plan rather than let agents stop. If you have a one-off need — backfilling enrichment across thousands of records, say — flag it to your account team first so we can plan for the run.
Rate limits
Section titled “Rate limits”A few limits keep usage steady across every workspace.
Agent runs are throttled per workspace. Agents run continuously across your whole book, but Sonora caps how many run at once for a given workspace. Day-to-day this is invisible; it only matters for bulk operations, which is another reason to flag large backfills to your account team ahead of time.
The API is rate-limited per access token. If you connect Sonora over MCP, each access token is limited to 60 requests per minute. Past that, calls return Rate limit exceeded. Retry after Xs. until the window resets. Spacing out requests or backing off on that message keeps a client well under the limit.