Compare gateways
Experiential vs Kong AI Gateway
Kong runs on your own provider keys, while Experiential is a hosted gateway with a catalog, credits and data controls that Kong can sit in front of.
Summary
Kong AI Gateway is a set of plugins (ai-proxy, ai-proxy-advanced and others) on Kong Gateway. It runs self-hosted or on Konnect. You bring every provider key, and Konnect bills per control plane, per model and per request. Experiential is a hosted gateway that also sells model access at list price. You can keep Kong as your edge and point its upstream at Experiential.
plugins:
- name: ai-proxy
config:
route_type: llm/v1/chat
auth:
header_name: Authorization
header_value: "{vault://env/openai-auth}"
model:
provider: openai
name: gpt-6-lunaSide by side
| Experiential | Kong AI Gateway | |
|---|---|---|
| Hosting | Hosted | Self-hosted, or Konnect (serverless, hybrid, dedicated) |
| APIs | Chat, Responses, Messages, Gemini generateContent, embeddings, batches | Chat, Responses, embeddings, files, batches, audio, image, video route types |
| Native Anthropic and Gemini | Every catalog model | Native formats pass through to their own provider only |
| Model access | 980 listed models on credits, or your keys | 17 providers, your keys |
| Pricing | List price on tokens, no markup | Plus from $25 a month; $100 a month per model beyond 5; 1M requests included, then $200 per 1M |
| Failover | Ordered waterfall; slow lanes demoted | Load balancing with failover (ai-proxy-advanced, an Enterprise plugin) |
| Zero data retention | Pro: org switch, per request, verdict per response | Not documented (depends on your provider contracts) |
| Spend limits | Per-key caps, budgets | Token budgets per consumer group |
| Content guardrails | Standard pack or your own classifier | Prompt guard, semantic prompt guard, PII sanitizer, AWS, Azure, GCP safety plugins |
| Response cache | None | Semantic cache (Enterprise plugin) |
| SSO | Not yet available | Enterprise |
Where Experiential is better
- No provider accounts to open. Credits reach 980 models, and failover between providers needs no extra plugin. See Routing and failover.
- Data controls are a switch, not a contract review. Zero-retention and no-training routing are enforced per org or per request. See Data controls.
- No per-model or per-request platform fees. See Plans and pricing.
When Kong may fit better
You already run Kong for API management, need self-hosting, or want a semantic cache or Kong's guardrail plugins in your own network. Keep Kong in front and use Experiential as its upstream.
Migration notes
Two options. Keep Kong and repoint each ai-proxy plugin (the YAML at the top of this page): clients, consumers and Kong policies stay the same. Or remove Kong from model traffic and point clients at https://api.experientiallabs.ai/v1.
| What | Kong | Experiential |
|---|---|---|
| Endpoint | Your Kong route | config.model.options.upstream_url: https://api.experientiallabs.ai/v1/chat/completions (/v1/responses for Responses) |
| Provider | config.model.provider | openai (the OpenAI wire format) |
| Auth | Provider key in config.auth | Authorization: Bearer YOUR_API_KEY, held whole in a Kong vault reference |
| Model ids | config.model.name | The catalog slug from GET /v1/models |
| Rate limits, prompt guard | Kong plugins | Keep them in Kong, or use per-key limits and guardrails |
Switch with an agent
Paste this into your coding agent. It inventories the project first and asks before anything it cannot map.
POST/v1/chat/completionscurl https://kong.example.com/ai/chat \
-H "apikey: $KONG_CONSUMER_KEY" \
-H "Content-Type: application/json" \
-d '{"messages": [{"role": "user", "content": "Hello"}]}'Sources, checked 2026-11-19
- Kong AI Gateway: providers, load balancing, semantic cache, guardrail plugins, observability, deployment options.
- ai-proxy plugin: route types, native formats, 17 providers, auth configuration.
- Kong pricing: Plus and Enterprise fees, Enterprise plugins, SSO on Enterprise.
- Experiential: /llms.txt, GET /api/models.