It lets agents query for meaningfully relevant code instead of relying only on keyword-based matches. You can try it here: https://wild-card.ai/deepcontext

We originally built this at Wildcard while working on background coding agents that take a ticket and return a PR. One of the persistent challenges was surfacing the right code context to steer the agent, especially for tickets that were less explicitly described.

Most coding agents today rely on ripgrep or keyword-based search to pull in files. This works fine on small vibe-coded projects, but breaks down on real-world repos with deep folder structures and inconsistent naming. We kept seeing two failure modes:

1. The query pulls in too many loosely matched files and overwhelms the context window. 2. The actual code snippets the agent needs are missed or buried.

DeepContext is our solution to this. The agent automatically generates semantic queries based on the user’s prompt, and calls DeepContext as a tool to retrieve relevant code chunks designed to fit in context. We currently support TypeScript and Python symbols.

We saw that existing AI coding tools like Claude Code and Codex CLI also rely on ripgrep, and thought this might be useful for devs using these tools too. So we pulled it out as a standalone MCP server and open-sourced it.

We’d love for you to try it out and share any feedback.

You can browse the code here: https://github.com/Wildcard-Official/deepcontext-mcp

And install the MCP here: https://wild-card.ai/deepcontext