Lexa

Highlights

• Built a local project index that can travel with the repository.
• Exposed code search, symbol lookup, outlines, and dependency tracing.
• Designed patch operations around line-based changes for agent workflows.

Context

AI coding agents are only as useful as the project context they can access. Raw text search helps, but larger repositories need a more structured way to understand symbols, files, dependencies, and edit targets.

Lexa is a local code intelligence layer designed for both humans and MCP-connected agents.

What I focused on

I built Lexa around a portable project graph so a repository can be indexed once and queried repeatedly for higher-signal code navigation.

The CLI and MCP server expose practical operations: search, symbol lookup, file outlines, dependency tracing, and line-based patch targets.

Outcome

Lexa turns repository context into a reusable local asset, making agent-assisted development less dependent on brittle ad hoc file reads.