Context Co.

Inspiration

The friction of the "Context-Switching Tax"—the mental toll developers pay when constantly jumping between VS Code, Jira, Slack, and Confluence just to understand a task. We wanted a unified "Cursor-like" experience that brings that knowledge directly into the editor.

What it does

ContextCo is a context-aware AI coding agent built as a VS Code extension. It autonomously explores the codebase and retrieves relevant information from connected tools (Slack, Jira, Confluence) to answer complex queries. It features a custom chat UI and a "Review & Apply" workflow that allows users to vet AI-generated code changes using the native diff view.

How we built it

We architected the solution as a VS Code extension that interfaces with a custom backend. The core logic leverages the Model Context Protocol (MCP) to standardize how the agent coordinates between the local file system and external APIs (like Slack and Jira), ensuring modular and scalable tool integration.

Challenges we ran into

Designing the agent architecture to reliably orchestrate multiple data sources without hallucinating or overwhelming the context window. Implementing the "Review & Apply" flow required deep integration with VS Code's native API to ensure diffs felt seamless rather than intrusive.

Accomplishments that we're proud of

-Successfully integrating a custom Chat UI directly into the IDE. -Bridging the gap between static code analysis and dynamic team knowledge (Slack/Jira). -Building a functional "Review & Apply" mechanism that respects the developer's final authority on code changes.

Built With

• fastapi
• gemini
• javascript
• mcp
• python
• typescript