Skip to content

Settings: network & limits

This page covers three tabs on Settings (/admin/settings): Network Access, Rate Limits, and Spend Caps. See Settings: governance for the other governance tabs.

A rule table controlling which hosts an agent’s web.fetch calls (and similar egress) may reach. Each rule has a scope (Tenant, Group, or User, with an optional scope value), an action (ALLOW, DENY, or ASK), a host pattern (an exact host, a *.suffix wildcard, or a bare * for everything), and an optional note. Add rule creates a new one; each row has its own Delete.

By default, with no rules at all, network access is allow-all - nothing is blocked. To build an actual allowlist, add a broad TENANT * DENY catch-all rule, then add specific ALLOW entries for the hosts you want reachable; the more specific ALLOW entries take precedence over the catch-all DENY.

A table of per-(agent, provider) request and token limits: requests per minute (RPM) and tokens per minute (TPM). Leaving agent or provider blank makes the policy a wildcard, matching whatever isn’t covered by a more specific policy. Leaving the limit itself blank means unlimited; setting it to 0 means deny-all for that agent/provider combination. The table renders these as ”∞” and “deny-all” respectively so they’re not mistaken for typos. Add Policy creates a new row; each row has its own Delete.

Monthly LLM spend caps, scoped at three levels:

  • Org - a single cap for the whole tenant, shown with the current period’s spend against it as a utilization bar and percentage. Set org cap / Clear manage it.
  • Per-user - pick or type a user (a datalist suggests existing members), enter a cap amount, and Set cap. A table below lists every user with a cap, their current spend, their cap, and percentage used, each with a Delete to remove that user’s cap.
  • Per-agent - the same pattern, scoped to individual agents instead of users.