AevrinAevrin MCP limits
What capability reads, runs and scans cost in credits, plus compute time, concurrency and per-run limits on Free and Pro.
Three things spend credits here: retrieving a capability with get_capability (1 credit),
running one with run_capability (10 credits), and starting a scan with scan_mcp_server
(50 credits, the same as any other scan). Searching is free and is never counted, so you can
always find out whether Aevrin has something before retrieving it, and so is reading a scan
report with get_scan_report. Following a capability's instructions is not counted either,
because Aevrin is not doing that work: your agent is, on your machine.
A run that Aevrin answers from a recent result costs the same 10 credits, and spends none of your monthly compute time. A capability Aevrin returns from its short-lived copy, rather than reading it again, still costs the same 1 credit.
Capability execution is not enabled yet, so no run is charged today. These are the limits that apply.
Per plan
| Free | Pro | |
|---|---|---|
| Credits per month | 1,500 | 50,000 |
A read (get_capability) | 1 credit | 1 credit |
A run (run_capability) | 10 credits | 10 credits |
A scan (scan_mcp_server) | 50 credits | 50 credits |
| Searches, scan reports | free, never counted | free, never counted |
| Compute time per month | 600 seconds | 20,000 seconds |
| Runs per minute | 5 | 20 |
| Runs at the same time | 1 | 2 |
A read is one get_capability: one capability's details, its instructions and its tools. It
costs once per call, wherever the call comes from, because every surface asks the same Aevrin
API. Retrieving the same capability again costs again.
Running a capability spends its 10 credits as well as the read's 1 that retrieved it. Those are two different things, not one charged twice: the read is the capability itself, and the run is Aevrin executing it on its own servers. A retrieval that comes back with no instructions and no tools is not counted, because you did not get anything.
Credits are one pool shared with your scans and code reviews, refilled on the same rolling monthly window counted from your signup date, and a one-time top-up adds 5,000 credits that never expire. See Concepts → Credits.
Per run
| Free | Pro | |
|---|---|---|
| Memory | 512 MB | 1 GB |
| Time | 30 seconds | 60 seconds |
| Output | 64 KB | 256 KB |
A capability can set lower limits for itself; your plan's limits are the most it ever gets. A run that goes past its time or memory limit is stopped and reported as failed.
When you reach a limit
- Credits used up. The retrieval, run or scan is refused, and the message says what it costs, how many credits are left and when the pool refills. Searching keeps working, because it is never counted. Upgrade, top up, or wait for the refill.
- Monthly compute time used up. The run is refused before anything starts, and the message says so and when it resets.
- Too many runs at once, or too many in a short time. The run is refused and can be retried in a moment.
A run that Aevrin refuses, or that could not start because of a problem on Aevrin's side, costs nothing.