coqu

Code Query — a service for answering questions about codebases using AI agents.

Users interact through a chat UI to ask questions about their code — architecture, logic, dependencies, bugs. The same capabilities are available through a REST API and MCP, so other agents and tools can query codebases programmatically.

Each query targets a specific project (codebase), agent (e.g. Claude Code), and mode (controls system prompt and response length — from tiny one-liners to deep exhaustive analysis).

Runs locally for development or deploys to a server via Docker. Can be exposed to the internet through Cloudflare Tunnel.

Why coqu

In large projects with microservice architecture and dozens of repositories, an agent working in one repo often needs context from others. coqu solves this: any agent can query other codebases through the REST API or MCP without leaving its own repository. This turns coqu into a shared knowledge layer across the entire project — agents in different repos can discover architecture decisions, understand dependencies, and coordinate through a single service.

Use case

Your company has two mobile apps — iOS and Android — in separate repositories. In-app purchases were implemented on iOS, and now the Android team needs to replicate the feature. The developer needs to understand the iOS implementation — architecture decisions, edge cases, error handling — without spending hours reading unfamiliar Swift code. With coqu, both repositories are connected, and any developer or AI agent can query the other platform's codebase to understand how a feature was built and replicate the same logic on their own platform.

Stack

• Frontend — React + Vite
• API — Express + Prisma
• Database — PostgreSQL 16
• Infrastructure — Docker, Cloudflare Tunnel
• Language — TypeScript (strict)
• Package manager — Yarn (workspaces)

Quick start

corepack enable
yarn install
cp .env.example .env
yarn dev

This starts PostgreSQL in Docker, then runs the API (:4000) and Web (:3000) locally.

Docker

# Local
docker compose up --build

# With Cloudflare Tunnel
docker compose --profile tunnel up --build

Documentation

📖 Online docs