V.A.L.A.A.R. — Personal AI Command Center

An open-source AI productivity platform combining persistent task management, a Streamable HTTP MCP server, and Amazon Bedrock integration to connect conversational AI with real actions.

Comment

Inspiration

I wanted to build more than a conventional chatbot. I envisioned a personal AI command center that connects conversational assistance with real, persistent actions. This idea led to V.A.L.A.A.R. — a platform designed to bring task management, AI-assisted workflows, and tool integrations together in one place.

What it does

V.A.L.A.A.R. combines a web dashboard, a FastAPI backend, PostgreSQL persistence, a Model Context Protocol (MCP) server, and an Amazon Bedrock integration.

Its MCP server exposes four task-management tools:

  • list_tasks — retrieve existing tasks.
  • create_task — create a task.
  • update_task_status — change a task's status.
  • delete_task — delete a task when explicitly requested.

The tools operate on persistent PostgreSQL task records shared with the application's backend. The MCP server uses Streamable HTTP transport at /mcp, allowing compatible MCP clients to interact with these task-management capabilities.

The project also includes an AI Assistant interface integrated with Amazon Bedrock through Boto3.

How we built it

We built the frontend using React, TypeScript, and Vite, with Python and FastAPI powering the backend. PostgreSQL stores task data, while SQLAlchemy manages database interactions. The MCP server is implemented using the Python MCP SDK and exposes task operations over Streamable HTTP. Amazon Bedrock integration uses the AWS SDK for Python (Boto3).

A central design goal was to let the dashboard, REST API, and MCP tools work with the same underlying task data instead of maintaining separate task stores.

Challenges we ran into

Bringing the frontend, backend, database, MCP server, and AI service together required coordinating multiple components and their configurations.

One significant challenge was getting Amazon Bedrock ready for live inference. We encountered account setup requirements and token-throttling errors during testing. This limited our ability to validate successful live AI responses, so we are careful to distinguish implemented integration from functionality successfully demonstrated end to end.

Accomplishments that we're proud of

We built a unified application that combines persistent task management with an MCP server exposing four practical tools. We are particularly proud of connecting the MCP tools to the same database used by the application's backend, making the tools capable of interacting with real task records rather than mock data.

We also established an integration path for Amazon Bedrock and created a web interface that brings the project's functionality together.

What we learned

We learned how MCP can expose application capabilities to compatible AI clients, how Streamable HTTP supports communication with an MCP server, and how a shared database can connect multiple interfaces to consistent application state.

We also gained experience integrating AWS services into a Python application and learned the importance of separating successful implementation from successful live-service validation.

What's next for V.A.L.A.A.R. — Personal AI Command Center

Our next priorities are to complete protocol compatibility validation, improve the reliability of the AI Assistant, strengthen setup and deployment documentation, and expand automated testing.

We also plan to improve the experience of connecting compatible MCP clients and eventually explore direct Alexa+ integration where supported. Production deployment, authentication hardening, and further reliability improvements remain future work.

V.A.L.A.A.R. is an open-source project under the MIT License and is actively being developed.

Built With

Share this project:

Updates

Submission history