An MCP server for the docs of any repo that uses Sphinx
Faster doc search for AI agents than vanilla Sphinx, with RRF hybrid retrieval.
Agent-optimized CLI design but competes with Cursor and Continue directly.
AI agent developers, terminal-based workflow users
Cursor · Continue · Sourcegraph Cody
GitHub: https://github.com/nozomio-labs/nia-cli
We originally built Nia as an MCP server and API, but a lot of the most capable agents today, like Claude Code, Cursor, and custom shell-based pipelines, live in the terminal. For those agents, a CLI felt like the most natural interface.
What it can do:
- Index repos, docs, papers, datasets, and local folders - Search across indexed sources with hybrid retrieval - Research codebases or personal data with cited outputs - Persist context so agents can save plans, decisions, and notes, then resume later without losing the thread - Search without indexing across the web, GitHub, and package registries
A few things we designed specifically for agents:
- structured JSON output - async-first workflows - non-interactive commands - composable multi-step pipelines - local folder sync and watch support
A few example commands:
nia repos index vercel/ai nia search query "How does auth middleware work?" nia oracle job "How does TurboPuffer handle vector quantization internally?" nia contexts save "Auth refactor plan" --content "$(cat plan.md)"
Install:
bunx nia-wizard@latest
Open source repo: https://github.com/nozomio-labs/nia-cli
Blog post: https://trynia.ai/blog/introducing-nia-cli
Would love feedback, especially from people building terminal-native agent workflows.
Faster doc search for AI agents than vanilla Sphinx, with RRF hybrid retrieval.
Chat-with-your-codebase tool, but Cursor, Continue, and Cody already own this.
Local doc indexing for Claude Code beats context-switching to browser tabs.
Surfaces architectural decisions from git history — Cursor and Continue don't do this.
Treats Markdown files as a database so fresh agents don't need to read 70k token chat logs.
Useful MDN index, but MDN's own search and navigation already cover this.