OpenAI-Compatible API Gateway for AWS Bedrock and AI Services
Deploy any OpenAI-compatible application on AWS Bedrock—no code changes required. Access 80+ models from Claude, Llama, Nova, and more with enterprise-grade privacy, compliance controls, and AWS direct pricing.
stdapi.ai is an OpenAI-compatible API gateway for AWS Bedrock and AI services. Drop-in replacement for OpenAI SDK—works with LangChain, Continue.dev, Open WebUI, n8n, and 1000+ tools. Deploy your existing OpenAI applications to AWS without code changes.
- 🔌 Production-Ready OpenAI API Compatibility – Full support for chat, embeddings, images, audio (speech/transcription/translation), and more. Applications designed for OpenAI's API work instantly.
- 🔒 Enterprise Compliance & Data Sovereignty – Configure allowed AWS regions to meet your compliance requirements. All inference stays in your AWS account—data never shared with model providers or used for training.
- 🌍 Multi-Region Bedrock Access – Automatic cross-region inference profile selection for optimal availability and pricing. Access models across multiple AWS regions through one unified endpoint.
- 💰 AWS Direct Pricing, No Markup – Pay-per-use pricing with no subscriptions. Pay only AWS Bedrock rates for exactly what you use—no monthly minimums or capacity commitments.
- ⚡ Advanced Model Capabilities – Reasoning modes (Claude 4.6+, Nova 2), prompt caching, guardrails, prompt routers, application inference profiles, and service tiers built-in.
- Anthropic – Claude 4.6+ with extended reasoning capabilities
- Amazon – Nova 2 family for cost-effective performance
- Meta – Llama 4 for open-source flexibility
- DeepSeek – v3.2 for advanced code generation
- OpenAI, Mistral AI, Google, Cohere, Stability AI, Qwen, Moonshot, Nvidia, and more
- Switch models instantly without code changes—no vendor lock-in
Unified under OpenAI API:
- Amazon Polly – Natural text-to-speech synthesis
- Amazon Transcribe – Accurate speech recognition with speaker diarization
- Amazon Translate – Multi-language translation support
- Chat completions with reasoning modes
- Image generation and editing (Stable Diffusion)
- Audio speech, transcription, and translation
- Embeddings for semantic search and RAG
- Complete AI workflows in one API
Build ChatGPT-like experiences with AWS Bedrock models and complete privacy control.
What you can build:
- Private team chat with Open WebUI or LibreChat
- Customer support assistant with RAG-enabled document search
- Internal knowledge base with multi-modal capabilities (text, voice, images)
Tools: Open WebUI, LibreChat, Chatbot UI
Integrate AWS Bedrock into business processes through visual workflow builders.
What you can automate:
- Customer support ticket classification and response generation
- Automated content creation for blogs, social media, email campaigns
- Document workflows with AI summarization, translation, and classification
- Data extraction, transformation, and analysis
Tools: n8n, Make (Integromat), Zapier
Enhance development with AI-powered coding assistants in your IDE.
What you can do:
- Real-time code completion in VS Code, JetBrains IDEs, Cursor, Windsurf
- Natural language to code generation with Claude and specialized coding models
- Chat with your codebase, explain functions, refactor code
- Build with LangChain, LlamaIndex, Haystack using AWS Bedrock
Tools: Continue.dev, Cline, Cursor, Windsurf, Aider, LangChain, LlamaIndex
Transform knowledge bases with AI-powered insights and semantic search.
What you can do:
- AI writing assistance to generate, edit, and improve content
- Semantic search to find notes by meaning, not just keywords
- Auto-summarization to extract key points from long documents
- Smart organization with automatic tagging and linking
Tools: Obsidian, Notion AI integrations, Logseq, Roam Research
Deploy intelligent AI assistants to Slack, Discord, Teams, and Telegram.
What you can build:
- Team Q&A bot for instant answers to common questions
- Documentation assistant that searches and cites internal docs
- Task automation via chat (create tickets, schedule meetings)
- Custom workflows with company-specific commands
Tools: Slack Bot, Discord Bot, Microsoft Teams Apps, Botpress
Build self-directed AI agents for complex multi-step tasks.
What you can build:
- Research agents for autonomous web research and analysis
- Multi-agent systems for collaborative problem-solving
- Self-improving workflows that adapt to results
- Autonomous development and testing systems
Tools: AutoGPT, BabyAGI, LangGraph, CrewAI, Semantic Kernel
- ✅ No code changes required – Just update the API endpoint in your application settings
- ✅ Access 80+ models – Claude 4.6+, Nova 2, Llama 4, DeepSeek v3.2, Stable Diffusion, and more
- ✅ Enterprise data control – All processing stays in your AWS account
- ✅ Pay-per-use pricing – No subscriptions, pay only AWS Bedrock rates for actual usage
- ✅ AWS-native features – Leverage prompt caching, reasoning modes, and guardrails through standard OpenAI API
📚 View Complete Use Cases & Integration Guides →
stdapi.ai is available on AWS Marketplace with commercial licensing, hardened containers, and streamlined deployment.
- ✅ Commercial License – Use in proprietary applications without AGPL obligations or source disclosure requirements
- ✅ Hardened Container Images – Security-optimized, regularly scanned for production workloads
- ✅ Regular Security Updates – Timely patches and vulnerability fixes to keep your deployment secure
- ✅ Terraform Deployment Module – Production-ready infrastructure following AWS Well-Architected Framework
- ✅ Enterprise Support – Professional support for deployment, configuration, and troubleshooting
- ✅ OpenTelemetry & Observability – Built-in monitoring and debugging capabilities
Community Edition: Free Docker image available for local development and testing.
1. Deploy to AWS in minutes Launch via Terraform module on ECS, or run the Docker image locally for development.
2. Point your application to stdapi.ai
Change only the base_url in your OpenAI client. All existing code, prompts, and workflows continue working.
3. Access AWS Bedrock models immediately Use Claude, Nova, Llama, or any Bedrock model. Switch between models, regions, and providers without changing application code.
Zero lock-in: Standard OpenAI API means you can switch back or to another provider anytime.
Deploy stdapi.ai to your AWS account in minutes using our Terraform module:
module "stdapi_ai" {
source = "stdapi-ai/stdapi-ai/aws"
version = "~> 1.0"
}Then make your first API call:
from openai import OpenAI
client = OpenAI(
api_key="YOUR_API_KEY",
base_url="https://YOUR_DEPLOYMENT_URL/v1"
)
response = client.chat.completions.create(
model="anthropic.claude-sonnet-4-5-20250929-v1:0",
messages=[{"role": "user", "content": "Hello from AWS!"}]
)
print(response.choices[0].message.content)- Python 3.14 or higher
- uv package manager
- AWS credentials configured
-
Clone the repository
git clone https://github.com/stdapi-ai/stdapi.ai.git cd stdapi.ai -
Install dependencies
uv sync --frozen --extra uvicorn
-
Login to AWS
# Login using AWS SSO aws sso login --profile your-profile-name # Or configure your default profile aws configure sso
-
Configure the application
# Core AWS Configuration (auto-detects current region if not set) export AWS_BEDROCK_REGIONS=us-east-1 # Optional: defaults to current AWS region # S3 Storage (required for certain features like image generation, audio) export AWS_S3_BUCKET=my-dev-bucket # Create bucket in same region as AWS_BEDROCK_REGIONS # Enable API documentation (helpful for development) export ENABLE_DOCS=true # Logging Configuration export LOG_REQUEST_PARAMS=true # Enable detailed request/response logging for debugging
-
Run locally
uv run uvicorn stdapi.main:app --host 0.0.0.0 --port 8000
-
Test the API
curl http://localhost:8000/v1/models
- Follow existing code style and conventions
- Add tests for new features
- Update documentation for user-facing changes
- Ensure all tests pass before submitting PR
- Official Documentation – Complete guides and API reference
- Getting Started – Deployment and configuration
- API Reference – Detailed API documentation
- Licensing – AGPL vs Commercial licensing
This project is dual-licensed:
- AGPL-3.0-or-later – Free for open-source projects that share alike
- Commercial License – Available via AWS Marketplace for proprietary applications
The AWS Marketplace version provides full commercial rights, no source disclosure requirements, and production-ready infrastructure.
We welcome contributions! Whether it's:
- 🐛 Bug reports and fixes
- ✨ New features and enhancements
- 📖 Documentation improvements
- 💡 Ideas and suggestions
Please feel free to open issues or submit pull requests.
- 🐛 Issues: GitHub Issue Tracker
- 📖 Documentation: stdapi.ai
- 💖 Sponsor: GitHub Sponsors – Support the project's development
Sponsorship benefits include priority support, feature prioritization, dedicated development time, SLA for critical issues, and influence on the project roadmap. View sponsorship tiers →
- 🌍 Multi-region Bedrock access – Automatic cross-region inference profile selection for optimal availability and pricing
- ⭐ Advanced model capabilities – Reasoning modes (Claude 4.6+, Nova 2), prompt caching, guardrails, service tiers
- 🔌 Complete API coverage – Chat, embeddings, image generation/editing, audio speech/transcription/translation
- 🎯 AWS AI services integration – Amazon Polly (TTS), Transcribe (STT with diarization), Translate—unified under OpenAI API
- 📊 Observability & debugging – OpenTelemetry, request/response logging, Swagger/ReDoc interfaces
- 🔒 Secure by default – API keys in Systems Manager, CORS controls, SSRF protection, hardened containers
Get Started with AWS AI
Documentation • AWS Marketplace • GitHub Issues
Made with ❤️ for the AWS and AI community