TokenGate
Start free
Privacy and cost control for AI apps

Ship AI features.
Meter usage, not people.

TokenGate protects provider keys, publishes your AI instructions, and enforces monthly allowances per installation. End-user requests and AI responses pass through without being stored.

Start building freeSee how data flows →
No conversation content persisted Random installation UUIDs only
tokengate.aceves.mx Secure
EXAMPLE APP

Usage overview

Last 30 days⌄
Total tokens84.2KMeasured usage
Estimated cost$1.86Clearly marked estimate
Installations42No user accounts required
Token usage Input Output
Jul 28Aug 4Aug 11Aug 18Aug 25
Cost per active user$0.065
↓ 14%
MCP connectionAgent connected
Route across the models your app needs
OpenAIAnthropicGeminiOpenRouterHugging Face+ custom

The short version

Conversation content is routed, not retained.

TokenGate stores the instructions your team publishes. It does not store end-user request content or AI responses. A random installation UUID becomes a keyed, project-scoped fingerprint before usage is persisted.

Verify the full data lifecycle

Keep a free feature sustainable

Give every installation an allowance.
Stop the next request when it is spent.

Usage limits protect a free or freemium app from unlimited AI calls. Cost figures are estimates; the allowance is the enforcement mechanism.

An allowance without an account

Meter a random app installation instead of asking for an email, username, phone number, or database ID.

  • Monthly installation limits
  • Block and forget controls
IDOne installationdevice_8f1...c9a
Active
August usage38,420 / 50,000 tokens
77% usedRenews in 7 days
Custom allowanceStop requests

Fix the prompt, not the release

Publish immutable prompt versions and choose what is live. Roll forward safely while your clients keep the same endpoint.

  • Server-side version control
  • Text and image inputs
meal-plannerProduction endpoint
Live
LIVE VERSIONv18Published 2h ago
System promptYou are a concise nutrition assistant. Use the user's dietary preferences and...
Model routing gpt-5-miniPrimary

Keep provider keys off devices

Encrypt credentials per organization and route app traffic through project-bound access keys.

See where usage goes

Inspect tokens, estimated spend, cost per installation, and monthly projections by project and prompt.

Choose the right model per prompt

Keep multiple providers ready and change model routing without touching client code.

Model Context Protocol

MCP first.
Dashboard optional.

Connect your coding agent to TokenGate and manage the platform where you already build. Create prompts, publish versions, rotate keys, inspect usage, and control installations through scoped developer credentials.

Project-scoped access Permission-level controls Built for agent workflows
Connect your project
AI coding agentConnected
›

Create a production prompt called meal-planner, use Gemini 2.5 Flash, and publish the first version.

TokenGate MCP

Creating prompt and publishing version...

Prompt deployed

/api/prompts/meal-planner · version 1 is live

›

One inspectable data path

Your app sends a random UUID.
TokenGate stores a fingerprint.

The request content passes through TokenGate to the AI provider selected by your team. TokenGate keeps only the operational metadata needed for allowances and reporting.

  1. 01
    Validate the installationThe identity header accepts only a random UUID, never an email or account ID.
  2. 02
    Enforce the allowanceTokenGate stops later requests once the monthly usage limit is spent.
  3. 03
    Route without retainingConversation content goes to the chosen AI provider and is not persisted by TokenGate.
RequestResponse•••
const result = await fetch("https://tokengate.aceves.mx/api/prompts/meal-planner",{method: "POST", headers: {Authorization: `Bearer ${APP_KEY}`,"X-Installation-Id": installationId}, body: JSON.stringify({ query: message })});
200 OK · 842 tokens · $0.0021

A boundary people can verify

Privacy does not stop at a slogan.

Our trust page lists what TokenGate receives, fingerprints, stores, forwards, and deletes. It also names the boundary: your chosen AI provider processes each request under its own data policy.

Read the public trust page →
90 daysDetailed usage retentionKeyedProject-scoped fingerprints0Conversation bodies stored

Built by an indie developer, for builders

Keep the feature useful.
Keep the spend bounded.

Start with one project, one provider key, and a monthly allowance for every installation.

Start building free