Skip to content
Now in early access

The AI gateway for
teams that ship

One API for every LLM provider. Built-in observability, access control, prompt versioning, regression testing, and model fallbacks — so you can focus on building, not plumbing.

gatelit · config
// Same call, any provider
const response = await gatelit.chat.completions.create({
messages: [{ role: "user", content: "Summarize this document" }],
model: "openai/gpt-4o",
});

Click a provider. Only the model field changes — nothing else.

15+

LLM providers supported

Any schema

Bring your own, we convert

<5ms

Added latency

Free

During early access

Features

Everything you need to run AI in production

Gatelit replaces a half-dozen internal tools with a single, consistent API layer.

Provider routing

Send requests to any model across OpenAI, Anthropic, Google, and 15+ providers through a single endpoint. Swap providers without touching your codebase.

Model fallbacks

Define fallback chains so requests automatically retry on alternate models or providers when the primary is unavailable or rate-limited. Zero-downtime failover.

Full observability

Every request logged: token counts, latency, cost, response content. Drill into individual calls, filter by user, track spend by provider. Export to your own tools.

Prompt versioning

Semantic versioning for prompts with full history, side-by-side diffs, and instant one-click rollback. Deploy changes without redeploying your backend.

Reusable partials

Extract shared prompt fragments — system messages, instructions, guardrails — into partials. Reference them across prompts. Edit once, update everywhere they're used.

Regression testing

Define test suites with input/output assertions. Run prompts against known inputs and validate responses before shipping. Catch regressions in CI before users do.

JSON schema validation

Constrain LLM outputs with JSON schemas. Validate responses against your type definitions and reject malformed output before it reaches your application code.

Access control

Issue API keys with granular permissions — per user, per team, per project. Track usage per key, set rate limits, and rotate credentials without downtime.

Caching & rate limiting

Cache identical LLM requests to save costs. Set per-user or per-endpoint rate limits to protect your provider accounts from runaway spend and abuse.

How it works

Three steps to get running

No migration, no rewrite, no downtime. Start with the free tier.

1

Install the SDK

Add the Gatelit SDK to your backend. It wraps the OpenAI-compatible interface, so existing code works immediately.

2

Configure your providers

Add API keys for OpenAI, Anthropic, Google, or any supported provider in the dashboard. No code changes needed.

3

Route and observe

Your app sends requests through Gatelit. Every call is logged, metered, and routed to the right model automatically.

Providers

Works with every major LLM provider

One integration, every model. Swap providers in a single config change.

OpenAI
Anthropic
Google Gemini
Mistral
Cohere
DeepSeek
Together AI
Groq
Fireworks

...and any OpenAI-compatible endpoint.

Pricing

Planned pricing

Free during beta. Final pricing will be usage-based. No surprises.

Hobby

Freeduring beta

For solo developers and side projects.


  • Up to 10,000 requests / month
  • 3 provider connections
  • 7-day log retention
  • Community support
  • Basic analytics
Join waitlist
Most popular

Pro

Coming soon

For growing teams shipping AI features.


  • Everything in Hobby, plus
  • Unlimited provider connections
  • 30-day log retention
  • Prompt versioning & playground
  • Regression testing & partials
  • Priority email support
Get notified

Enterprise

Custom

For organizations with scale and compliance needs.


  • Everything in Pro, plus
  • Unlimited requests
  • SSO / SAML
  • Custom log retention
  • Dedicated support & SLA
Talk to us

FAQ

Frequently asked questions

Do I need to change my existing code?

No. Gatelit is an OpenAI-compatible proxy. If you use the OpenAI SDK today, you change the base URL to point at Gatelit and add an API key — the rest of your code stays the same. Migration takes under a minute.

What providers do you support?

OpenAI (all models including GPT-4o, o3, o4-mini), Anthropic (Claude Opus, Sonnet, Haiku), Google (Gemini), Groq, Together AI, Fireworks, DeepSeek, Mistral, Cohere, and any OpenAI-compatible endpoint. We add new providers regularly — reach out if you need one.

How does model fallback work?

You define fallback chains in the dashboard: 'try gpt-4o first, fall back to claude-sonnet-4 if it's rate-limited, then to gemini-2.5-flash as a last resort.' The gateway handles retries and failover transparently. Your users never notice an outage.

Is my data stored?

You control log retention. By default we store request/response metadata (tokens, latency, cost) for analytics. Response content logging is opt-in per API key. Enterprise plans support custom retention policies and on-premise deployments where your data never leaves your infrastructure.

How does billing work?

Gatelit charges based on request volume and log retention. You still pay your LLM providers directly — we don't mark up model costs. You bring your own API keys. No credit card required on the free tier.

How is this different from Helicone, Portkey, or Langfuse?

Gatelit combines observability, gateway routing, and prompt management in one layer — plus regression testing, JSON schema validation, and reusable partials. Helicone focuses on logging, Portkey on routing, Langfuse on tracing. We do all three plus prompt engineering workflows in a single integration.

Can I self-host?

Not yet. We're focused on making the hosted product excellent during early access. Self-hosting and on-premise deployment are on the roadmap for Enterprise customers — reach out if that's a requirement for you.

Waitlist

Join the waitlist

We're opening access in small batches. No spam, just updates when your spot is ready.

What do you use for prompt management today?

Ready to stop plumbing and start shipping?

Start with the free tier. No credit card required. Get your first request through the gateway in under two minutes.