TL;DR: I built an open source app that runs 100% locally, it provides a TUI to index repos, private or public and make them searchable via an MCP. The primary use case for me was to be able to quickly reference the libraries I care for, such as effect-ts, ai-sdk and others. I didn't want to just reference documentation, I wanted to allow my coding agents to search through their code and learn deeply the best patterns of usage and what sits behind their APIs.
One can achieve similar results by allowing the agent to search directly in the file-system. However, it is clunky, the discoverability of repos insn't good and the agent is limited to 'grep'.
WIth repobase, the agents can discover all the repos you have indexed, it can search them semantically (e.g "How do I do image generation using ai-sdk using grok as the provider?").
To use it, you just need to
1. 'npm install -g repobase'
2. run 'repobase' to access the TUI and view and index repos
3. Add the MCP server to Cursor, Claude Code or any other tool you use for coding/chatting.
Posting here to see if I find other with a similar pain that could benefit from the app. I'd love to explore how to evolve this tool as I see it could be also beneficial for teams working with many micro-frontends or micro-services. Where unifying the context and discoverability of repos could be valuable.
Looking forward to any ideas, suggestions or folks just wanting to give it a go :)
A local first context engine for Cursor, Claude Code and more · HackerTrans
WIth repobase, the agents can discover all the repos you have indexed, it can search them semantically (e.g "How do I do image generation using ai-sdk using grok as the provider?").
To use it, you just need to
1. 'npm install -g repobase' 2. run 'repobase' to access the TUI and view and index repos 3. Add the MCP server to Cursor, Claude Code or any other tool you use for coding/chatting.
Posting here to see if I find other with a similar pain that could benefit from the app. I'd love to explore how to evolve this tool as I see it could be also beneficial for teams working with many micro-frontends or micro-services. Where unifying the context and discoverability of repos could be valuable.
Looking forward to any ideas, suggestions or folks just wanting to give it a go :)