Limits and what is included
One reference for every limit ax1om enforces, per plan. Prices and the tier comparison live on the pricing page; this page is the operator’s copy of the numbers.
Two kinds of limit appear below. A soft cap keeps working past the line and tells you where you are. A hard limit refuses. Which is which is stated per row.
Training and building
Section titled “Training and building”| Limit | Free | Pro | Enterprise |
|---|---|---|---|
| Training records | 250K | 500K | 1.5M · scales |
| Active models (soft cap) | 10 | 50 | 200 |
| Runs per day | 5 | 25 | 100 |
| Runs per week | 15 | 100 | 500 |
| Training runs included per month, agent usage billing* | 5 | 50 | 150 |
| Timing refreshes included per month, agent usage billing* | None | 20 | 60 |
| CRM connections | 1 | 3 | 5 |
Insights come with every plan, including Free: feature importance, the Feature Stability Score, score distribution, the lift curve, field diagnostics, and the assistant.
* Applies when agent usage billing is enabled for your workspace; it is off by default. These allotments count agent-initiated work and human-initiated work out of one pool per meter. Rates for use past an allotment are published with the agent usage billing documentation, not here.
Delivery
Section titled “Delivery”| Limit | Free | Pro | Enterprise |
|---|---|---|---|
| Score export (CSV) | Not included | Included | Included |
| CRM writeback | Not included | Included | Included |
| Score maintenance | Not included | Weekly · all records | Weekly · all records |
| Records scored over the API | No live API | 10K/mo | 30K/mo included · scales |
| API keys | Not included | 5 | 100 |
Per-request limits
Section titled “Per-request limits”These apply on every plan that has the live API.
| Limit | Value | What happens at the line |
|---|---|---|
| Requests per second, per key | 10 | rate_limited · slow down and retry |
| Records per batch call | 1000 | record_limit_exceeded · split the batch |
CSV upload size (MCP upload_csv) |
100MB, .csv only |
The upload is refused |
Batch up to 1000 records per call to stay well under the rate limit.
The monthly scoring meter
Section titled “The monthly scoring meter”The monthly allocation for records scored over the API is a soft cap:
- Inside your allocation, requests score normally.
- Past it, live keys on a paid plan keep scoring in a grace band up to 2x the limit. The response carries
in_grace: trueandremaininggoes negative. - At the backstop, scoring pauses with
quota_exceededuntil the month resets or you upgrade.
Test keys (ax1m_sk_test_) meter into their own free bucket with a hard cap, so integration work never touches your live allocation.
There are no usage overages on any plan. Passing a limit changes what the API returns; it never changes what you are charged. See billing.
What you see at a limit
Section titled “What you see at a limit”| You hit | The response |
|---|---|
| Monthly scoring allocation | in_grace: true, then quota_exceeded at the backstop |
| Per-key rate limit | rate_limited |
| Records per batch | record_limit_exceeded |
| CRM connection limit | The connection is refused when you authorize it |
Every one of these is a named code with prose behind it. See errors for the full table, and troubleshooting for what to do about each.
Enterprise limits are contract terms rather than fixed lines. Record volume, connected CRMs, and modules are quoted per org.
- How usage accrues · which actions move which meter
- Rate limits and usage · the
usageobject in every batch response - Errors · every code the API can return
- Pricing · plans, prices, and the full tier comparison