Provider-neutral routing
Publish stable model aliases, route them to configured providers, and define ordered fallback targets.
Route OpenAI-compatible inference, issue scoped API keys, enforce security policies, track usage, and govern MCP access from one gateway.
curl -fsSL https://install.tuenel.com | shfrom openai import OpenAI
client = OpenAI(
api_key="YOUR_TUENEL_API_KEY",
base_url="http://localhost:4060/v1",
)
response = client.chat.completions.create(
model="your-route-alias",
messages=[{"role": "user", "content": "Hello from Tuenel"}],
)Configure a provider and model route in the console before sending inference. Tuenel is a gateway, not a model host.
Gateway capabilities
One control plane for routing, credentials, policy, usage, and MCP operations.
Publish stable model aliases, route them to configured providers, and define ordered fallback targets.
Authenticate browser users and applications with organization, project, role, and API-key boundaries.
Inspect prompts, redact configured sensitive patterns, record findings, and apply warn, block, or incident actions.
Track token usage, configure model pricing, and apply quotas and budget rules across projects.
Register MCP servers, filter tool discovery, classify risk, and require approval where configured.
Provider credentials stay server-side while applications use scoped Tuenel API keys and stable route aliases.
OpenAI-compatible client
Authentication, quota, and policy
Aliases, targets, and fallback
PostgreSQL plus Redis counters
Frequently asked
Yes. The Apache-2.0 Community edition runs independently and does not require a Tuenel account, license, or connection to a Tuenel-operated service.
Managed cloud
Create an organization, configure a provider, publish a route, and send your first request without operating the gateway stack yourself.
Self-hosted Community
Clone the public Apache-2.0 repository and run the supported interactive installer.
curl -fsSL https://install.tuenel.com | sh