AI governance
your developers actually trust.

Multi-provider proxy with cost tracking, budget controls, and policy enforcement. Track what AI costs, not what it says.

Get early access View on GitHub

Metadata only by default. No prompts. No responses.

terminal

$ export ANTHROPIC_BASE_URL=https://proxy.yourco.com/v1
$ export OPENAI_BASE_URL=https://proxy.yourco.com/v1

# That's it. Your tools work exactly the same. ▮

Team spend this month+12%

$2,847.32

Claude

68%

GPT-4

24%

Gemini

Budget$2,847 / $5,000

Built for modern engineering teams

Open source

Self-hosted

Multi-provider

Privacy-first

Three steps. Full visibility.

Works with your existing tools and workflows.

Point

Route your AI tools through Alwakeel. One environment variable. No SDK changes.

export OPENAI_BASE_URL=https://proxy.yourco.com/v1

See

Real-time cost and usage dashboards. Know exactly what AI costs per team, project, and model.

$2,847 across 3 teams this month

Control

Set budgets, enforce policies, manage access. All without reading a single prompt.

57% of monthly budget used - 0 policy violations

Everything you need to govern AI usage.

Built for teams that move fast and need to stay in control.

Multi-provider proxy

One proxy for Anthropic, OpenAI, Gemini, Mistral, and Ollama. Route traffic through a single endpoint with zero vendor lock-in.

Cost intelligence

Real-time spend tracking per team, user, model, and project. Know where every dollar goes before the invoice arrives.

Budget controls

Spending limits at org, team, or user level. Automatic enforcement before requests are proxied. No surprises.

Policy engine

Model allowlists, token limits, rate limits. Sub-5ms policy evaluation in Rust. Define rules once, enforce everywhere.

Team management

Org, team, and user hierarchy with scoped API keys and role-based access control.

Privacy-first architecture

Usage records contain cost metadata only. Prompts and responses never touch our storage layer. This is architectural, not configurable - the fields do not exist.

Privacy is architecture, not a toggle.

Alwakeel tracks costs and usage patterns. It never stores what developers actually type or receive.

What your company sees

Total spend per team$2,847.32

Model usage breakdownClaude 68%

Token count1.2M tokens

Budget utilization57% of $5,000

Request count3,847 requests

What stays private

Prompt content

AI responses

Code snippets

Conversation context

Session history

The fields don't exist. UsageRecord has no columns for prompt or response content. This is architectural, not configurable. Developers can verify this in the source code.

Built for every role in engineering.

Zero friction. Full transparency.

Set one environment variable and keep using your tools. No SDK changes, no intercepted prompts, no surveillance. You can verify exactly what Alwakeel tracks by reading the source.

One env var to configure

No SDK or code changes

Verify tracking scope in source code

Same tools, same workflow

Works with every major AI provider.

Add new providers without changing client configuration.

Anthropic

Claude models

OpenAI

GPT models

Google

Gemini models

Mistral

Mistral models

Ollama

Local models

Open source

Self-host or use our cloud.

Deploy on your own infrastructure in minutes. Full control over your data, your policies, your costs.

Docker Compose

Single command to start. Perfect for evaluation and small teams.

Kubernetes

Production-ready deployment. Horizontal scaling built in.

Source available

Audit the code. Verify the privacy claims. Contribute upstream.

$ git clone github.com/alwakeel-ai/alwakeel
$ docker compose up

# API on :8080, dashboard on :3000

Start governing AI spend
without breaking developer trust.

Join the early access program and help shape the future of AI governance.

Free for teams under 10 developers during early access.

AI governance your developers actually trust.