No Vendor Lock-in
Stay free to choose — and swap — the best-fit model for each task. No multi-year contracts and no large up-front CapEx; pay for what you use and change course as the field moves.
Introduction
A production-ready, multi-agent conversational AI platform built on AWS Bedrock AgentCore and Strands Agents — for teams who want to run their own AI assistant on infrastructure they control, instead of renting per-seat access to someone else’s.
What you get
Section titled “What you get”The platform ships the pieces of a modern conversational AI product, wired together and ready to deploy:
- Streaming chat — token-by-token responses over SSE, with conversation history and automatic context management.
- Multi-agent tools — agents call tools across several protocols: direct functions, the AWS SDK, MCP servers, and remote A2A agents.
- RAG — ground answers in your own documents and knowledge bases.
- MCP Apps — rich, interactive tool UIs rendered inline in the conversation.
- Artifacts — generated documents and code rendered in a dedicated side panel.
- Voice — speak to the assistant and hear responses back.
Why an Internal AI Platform?
Section titled “Why an Internal AI Platform?”Commercial AI subscriptions are convenient, but at scale they lock you into someone else’s roadmap, pricing, and data practices. Running your own platform on infrastructure you control changes that calculus.
Data Sovereignty
Your data stays inside your own AWS account. Nothing leaves the security boundary you control.
No Training on Your Data
Models are used for inference only. Your prompts and data are never used to train third-party models.
No Per Seat Licensing
Per-seat pricing doesn’t scale — costs balloon with every user you add. A consumption-based, self-hosted platform means you pay for actual usage, not headcount.
Equitable Access
Everyone gets the same capable models — no tiered “haves and have-nots” based on who can afford a premium subscription.
Systems Integration
Teams can bring their own MCP servers to connect the systems they rely on, so the assistant can safely act on real data — all within your own controls.
Next steps
Section titled “Next steps” Quick Start Get the platform running locally in a few commands.
Architecture Overview How the agents, APIs, and infrastructure fit together.
Local Development Run the backend and SPA locally against a deployed environment.
Deployment Ship to AWS: platform stack, backend images, and the frontend.