LLM & MCP Gateway

The Gateway for LLM and MCP Traffic

Add rate limits, authentication, and LLM routing to any endpoint. Define policies in code and deploy through git push.

12 Pilot teams
50+ LLM API routes protected
3 days Time to first secured route
Internal early-access beta data, 2025-2026

One config file. Control AI traffic.

Your routes.json defines every gateway policy. No UI clicks, no drift across environments.

Your API any endpoint routes.json Zuplo Gateway auth-key rate-limit retry inbound policy chain LLM / MCP any provider

Each policy is a versioned JSON declaration. Commit routes, push, and deploy your gateway. No vendor lock-in or config drift.

What an AI gateway needs out of the box

Built for LLM traffic patterns, with MCP support alongside standard REST.

API Key Management

Issue, scope, and revoke consumer keys. Rotate without downtime. Tie every key to an auditable rate-limit policy.

Token-Aware Rate Limits

Limit requests by count or token budget. Use sliding, fixed, or concurrent limits, composed in config.

LLM Routing

Route across providers by cost, latency, or custom logic. Fallback chains keep your API online during provider outages.

MCP Gateway Support

Handle MCP sessions alongside standard REST traffic. One gateway and config for both protocol families.

Your gateway lives in Git

No click-ops or portal drift. Policy changes are pull requests, and deployments are traceable.

routes.json
{ "routes": [ { "label": "OpenAI proxy", "path": "/v1/chat/completions", "methods": ["POST"], "policies": { "inbound": ["api-key-auth", "token-budget-1m", "rate-limit-tier"] } }, { "label": "Anthropic fallback", "path": "/v1/messages", "methods": ["POST"], "policies": { "inbound": ["api-key-auth", "cost-attribution"] } }, { "label": "MCP tool server", "path": "/mcp/:session", "methods": ["GET", "POST"], "policies": { "inbound": ["jwt-auth", "mcp-session-guard"] } }, { "label": "Internal admin", "path": "/admin/*", "methods": ["GET", "POST", "DELETE"], "policies": { "inbound": ["saml-sso", "audit-log"] } } ] }

Deploy changes with: git push origin main

Teams shipping with Zuplo

LLM API rate limiting is hard to get right: token buckets, sliding windows, consumer quotas, and token-budget edge cases. Zuplo handles it declaratively, so I stopped writing custom middleware for it.

Kioni Mwangi API platform engineer, AI infrastructure company

We evaluated three gateways for our MCP product. Zuplo was the only one treating MCP as a first-class concept. Its session-guard policy saved us about two sprints of custom proxy work.

Rajiv Subramaniam Engineering lead at a developer tools startup

Simple, clear pricing

Start free. Upgrade as traffic grows.

Starter
Free
100K calls/month
Builder
$79/mo
1M calls/month
Scale
$299/mo
10M calls/month
Enterprise
Custom
Unlimited

Make your AI gateway a config file.

Start free. Add policies as you scale. No infrastructure to manage.