GATEWAY / CONTROL ROOM
Gateway ready
ONE ENDPOINT · EVERY PROVIDER · COMPLETE CONTROL

Your AI traffic,
under control.

A hands-on command center for Cloudflare AI Gateway. Send real requests through your custom hostname—or explore every control in a zero-key simulation.

https://housedoor.buggy-refreeze.uk
Launch console ↗
20+AI providers
1unified endpoint
∞routing patterns
<1mscontrol overhead*
*Illustrative demo metric, not a measured guarantee.
01LIVE REQUEST LAB

One request.
Every control.

Compose an OpenAI-compatible request, inspect the exact headers, and see the gateway response—including cache and trace metadata.

REQUESTSIMULATION
Connect credentials optional

Tokens stay in this tab’s memory and are never saved. Use BYOK or Unified Billing to avoid entering a provider token.

PER-REQUEST CONTROLS
Cache responseSet a request-specific TTL
Skip cacheForce a fresh provider response
Collect logsRequest and response observability
Stream responseRender tokens as they arrive
Require provider keyDisable Unified Billing fallback
RESPONSE— ms

Ready when you are.

Your request trace will appear here.
GENERATED REQUEST

        

      

    


    

      

        
02DYNAMIC ROUTING
Traffic that
thinks ahead.

        
Model routing lives at the gateway, not in application code. Segment users, split traffic, enforce budgets, and fall back when a provider fails.

      

      

        
dynamic/support-pilotVersion 4 · deployed

        

          
01Incoming requestRead metadata

          
02plan = “pro”?Conditional

          

            
YESOpenAIgpt-5.2

            
NOWorkers AIllama-3.3-70b

          

          
04Budget check$25 / team / day

          
05Responsewith trace

        

        
 Active path Alternative Healthy

      

    


    

      

        
03FEATURE MATRIX
Control every
dimension.

        
The request lab exposes request-level controls. Gateway-level features below are configured in Cloudflare and then applied consistently across providers.

      

      

        
◫PERFORMANCE
Semantic & exact caching
Serve repeat requests from the edge with configurable TTLs and custom keys.

        
⑂RELIABILITY
Retries & fallbacks
Recover from rate limits and provider failures with controlled retry strategies.

        
◎COST
Budgets & spend limits
Enforce dollar budgets by provider, model, user, team, or application.

        
◇SAFETY
Guardrails & DLP
Inspect prompts and responses for unsafe content and sensitive data.

        
⌘SECURITY
BYOK & authentication
Centralize encrypted provider keys and protect the gateway with authentication.

        
↝OPTIMIZATION
Auto Router
Select a cost-effective model for each request while maintaining response quality.

      

      

    


    

      

        
04OBSERVABILITY
See the whole
AI estate.

        
Use logs, analytics, cost attribution, user insights, classifications, custom metadata, and OpenTelemetry to understand every request.

      

      

        
Gateway pulseSimulated telemetry · last 24 hours

        
REQUESTS128.4K↑ 12.8%
P95 LATENCY842ms↓ 9.4%
CACHE HIT RATE68.7%↑ 5.1%
EST. COST$184.20↓ 16.2%

        

          
4K3K2K1K0

          
            
            
            
          
          
00:0004:0008:0012:0016:0020:00NOW

        

        
RequestsCache hitsDemo data

      

    


    

      READY TO TEST THE EDGE?
      
Point your app at
one smarter endpoint.

      
Your custom gateway hostname is already wired into the request lab.

      Open the console ↗