Now in beta — join the waitlist

The middleware layer
for production LLM apps

One SDK between your app and any LLM. Privacy, prompt management, observability, and evals — without the glue code.

View on GitHub →
integration.py
# Before: your app talks directly to OpenAI response = openai.chat(model="gpt-4o", messages=[...]) # After: one line change, full control import gateforge gateforge.init(api_key="sk-live-...") response = gateforge.chat(model="gpt-4o", messages=[...]) # ✓ PII stripped before leaving your system # ✓ Prompt versioned and tracked # ✓ Cost and latency logged # ✓ Fallback to Claude if OpenAI fails

Why Gateforge?

LLMs in production are harder than they look

🔒
Privacy risk
Every API call sends raw user data to OpenAI. Names, emails, medical records — all of it. One integration, one liability.
📊
Zero visibility
You don't know what each call costs, how long it takes, or when something breaks. Your bill arrives before your metrics do.
🔧
Prompt chaos
Prompts hardcoded across the codebase. Every change is a deploy. No versioning, no rollback, no A/B testing.

Modules

Everything you need, nothing you don't

Modular by design. Use what you need today, add the rest when you're ready.

PII Shield Available now

Anonymize sensitive data before it reaches any LLM. Restore it transparently in the response. Healthcare, finance, and legal profiles out of the box.

View open source →
Prompt Studio Coming soon

Version, test, and deploy prompts without code changes. A/B test variants across traffic. Roll back to any previous version in one click.

Observe Coming soon

Real-time cost tracking per request, per model, per user. Latency percentiles. Spend alerts before you get a surprise bill at the end of the month.

Eval Coming soon

Define test cases in plain language. Run them automatically when a prompt changes. Catch regressions before they reach production users.

How it works

One SDK, full control

Your App
→
Gateforge SDK
→
PII Shield
→
Prompt Studio
→
OpenAI / Anthropic / Mistral
↑ request path

Your Users
←
Rehydrated response
←
Observe + Eval
←
LLM response
↑ response path
Install in minutes
One line change to your existing integration. No infrastructure to manage, no new services to deploy. Works with your current LLM setup.
Provider agnostic
OpenAI, Anthropic, Mistral, local models. Switch providers with one parameter. Automatic fallback when a provider goes down.
Self-hostable
Deploy on your own infrastructure. Your data never leaves your system. Full audit logs. SOC 2 compliant by design.

PII Shield

Start with privacy. The rest follows.

PII Shield is open source and available today. The full Gateforge platform is coming soon.

Input from your app
"Patient Ana García, DNI 12345678A,
diagnosed with hypertension."
↓ PII Shield — healthcare profile
Output to LLM
"Patient [PERSON_001], [REDACTED],
diagnosed with hypertension."
↓ LLM response rehydrated
Final response to user
"Ana García should schedule a follow-up
appointment for her hypertension."
55+
Languages supported
7
Detection backends
6
Disposition actions
< 1ms
Regex mode latency
Try PII Shield free →

Early access

Be the first to know when we launch

Join the waitlist. Get early access, updates, and a lifetime discount for early adopters.

No spam. Unsubscribe anytime.

You're on the list! We'll be in touch soon.