Policy and spend control at the edge.

Rate Limits & Quotas

Per route, per key, per tenant — from JWT claims or headers.
One policy, enforced consistently at the edge.

Budgets

Cost-aware spend caps per key, team, or endpoint.
Throttle or cut off traffic when real usage crosses defined thresholds.

Noisy Neighbor Isolation

One tenant cannot degrade others.
Hard isolation via per-tenant limits and budgets.

Abuse Controls

Bot and crawler detection (ASN, fingerprinting signals).
Edge enforcement only — not a WAF, no payload inspection.

Versioned policies

Full version history and safe rollback. Policy changes are reviewed and deployed — not hidden in application code.

{
  "selector": { "path_prefix": "/v1/chat/" },
  "limit_keys": ["jwt:org_id"],
  "rules": [
    { "limit": "pro-tier" },
    { "budget": 500, "period": "1d",
      "unit": "USD" }
  ]
}

Shadow mode

Limits are calculated but not enforced. See exactly which requests would be blocked before switching to hard enforcement.

fairvisor validate ./policies.json
fairvisor test --dry-run
# logs what would be enforced
# on your real traffic

Edge enforcement

Sub-millisecond, in-process enforcement. If the control plane is unreachable, last-known policy remains active.

docker run -d fairvisor/edge \
  --config ./policies.json \
  --upstream http://your-api:8080

Public APIs with paid tiers

You enforce Free / Pro / Enterprise tiers.
Hard limits at the edge — not best-effort checks in code.

Partner integrations

Each partner has its own contract.
Per-integration limits and kill switches enforced in real time.

Outgrown your API gateway's rate limiting

High-volume APIs where per-request gateway pricing no longer makes sense.
Keep your gateway. Add a dedicated enforcement layer.

Predictable failure

Deterministic enforcement: warn → throttle → reject.
No surprise overage at month end.

LLM spend control — no SDK changes

Point your existing LLM clients at Fairvisor as a drop-in endpoint. Change the base URL and API key format — no application code changes. Policy and spend control apply transparently. See AI enforcement →

Multi-tenant SaaS

One tenant's traffic spike shouldn't degrade service for others.
Hard per-tenant isolation enforced at the edge — not patched in application code.