Super ProxySuper Proxy
Open source · Repository coming soon

AI access management for small teams

AI subscriptions are cheap. Managing them is chaos.

Give every teammate their own API key — only the models admins approve, with usage limits and full visibility.

Opens a pre-filled email to request early access. No newsletters.

See how it works

How it works

One gateway.
Access for every teammate.

01

Connect your providers

Admins connect the AI providers and API accounts the company is authorized to use.

Your providers

OpenAI Connected
Anthropic Connected
GLM Connected
Kimi Connected
Gemini Connected
02

Create individual API keys

Every teammate receives their own key. No provider passwords, shared accounts, or repeated logins.

Team members

AlexAlexDeveloperkey_alx_7f2...
MayaMayaDesignerkey_mya_9c1...
SamSamProductkey_sam_a3b...
JordanJordanOpskey_jrd_2d8...
AngelaAngelaMarketingkey_ang_5e4...
03

Approve models and limits

Admins control model access and usage limits for each employee, role, or team.

Key details – Alex

Active

Allowed models

GPT-4oAllowed
Claude 3.5 SonnetAllowed
GLM-4-PlusAllowed
Kimi K2Not allowed

Spend limit

$200 / monthEdit

Before vs after

From shared credentials to controlled AI access.

Replace provider passwords, repeated logins, and uncontrolled model access with individual keys managed from one place.

Before

  • Shared passwords and 2FA codes
  • Access to every provider model
  • One employee can consume the limit
  • No visibility by employee
  • Change passwords for everyone
  • Employees switch between providers

After

  • Individual API key per employee
  • Only admin-approved models
  • Per-user limits and controls
  • Usage tracked by employee
  • Revoke or rotate one key instantly
  • One OpenAI-compatible endpoint
Controlled access·Usage visibility·One key per teammate

Admin access, limits, and usage controls

Control access. Set limits.See everything.

Give each teammate the right models and limits. Track usage in real time and keep costs under control.

Manage model access

Choose which models each teammate can use.

AlexAlexDeveloper
GPTClaudeGLMKimi
SamSamProduct
GPTClaudeGLMKimi
MayaMayaDesign
GPTClaudeGLMKimi
JordanJordanOps
GPTClaudeGLMKimi
View all members

Granular access

Control every model for every user.

Smart limits

Prevent overuse and stay on budget.

Full visibility

Know who used what, and when.

Instant actions

Revoke, change, or update in one click.

Custom models & intelligent routing

Build the models your team actually needs.

One simple model name. Your gateway picks the best provider by task, cost, and remaining limits — with automatic fallback.

Employee request

POST /v1/chat/completions

{
  model: "company-fast",
  prompt: "Summarize today's support tickets."
}

Cheapest capable model. Avoids premium.

Gateway routes to

OpenAI72%
Anthropic41%
GLM87%Chosen
Kimi22%Fallback

GLM is cheapest with 87% of its limit left — ideal for routine work.

1Understand the request

Task type, quality bar, and latency preference.

2Evaluate in real time

Remaining limits, cost, performance, availability.

3Route & fallback

Pick the best model — fall back if a provider is down.

Open source trust

Your AI gateway should not be a black box.

Inspect the code, understand how requests are handled, and run the gateway in infrastructure you control.

Opens a pre-filled email to request early access. No newsletters.

src/routing/router.tsCodeBlame
1export async function routeRequest(req: Request) {
2  const context = await analyzeRequest(req);
3  const candidates = await getAllowedModels(req.user);
4  const scored = await scoreModels(candidates, {
5    task: context.task,
6    cost: context.costPriority,
7    limits: await getRemainingLimits(req.user),
8    latency: context.latencyPriority
9  });
10  const choice = selectBest(scored);
11 
12  try {
13    return await forward(req, choice);
14  } catch (err) {
15    const fallback = await getFallback(choice);
16    return await forward(req, fallback);
17  }
18}
Open-source core
Self-host it
Transparent routing
No credential exposure
Community first

Security by design

Built for control and privacy.

You decide what data is stored, where it goes, and who can access it. Every request passes through your gateway before it ever reaches a provider.

  • Minimal data retention
  • Local logs and full auditability
  • Role-based access control
  • Revoke access in one click
  • Designed for compliance
See all security features

FAQ

Frequently asked questions.

Everything you need to know about Super Proxy and how it helps your team use AI with the right access and controls.

Still have questions?

We're here to help.

Contact us

Yes. Every teammate gets an individual key, so access, limits, usage, revocation, and audits stay tied to the right person.

Yes. Admins can allow or block models per key, role, or team and update access without sharing provider credentials.

Yes. Point any OpenAI-compatible client at the Super Proxy base URL and continue using the same SDK and request format.

Super Proxy connects the AI providers and API accounts your company is authorized to use, then manages team access through individual keys.

The gateway is being built with self-hosting in mind. Join the waitlist to receive the repository and deployment instructions when they are published.

Open-source Super Proxy repository illustration

Open by design. Built for teams.

Super Proxy is being built in the open. Join the waitlist to get the repository, docs, and launch update when they are ready.

Join the waitlist