Showcases

Real examples of using Chub with Claude Code, Cursor, and other AI coding agents.

Express + OpenAI API project

Set up a Node.js project with auto-detected dependencies, pinned docs, team annotations, and MCP integration — in under a minute.

1. Create project and auto-detect deps

# Start with a standard package.json
cat package.json

{
  "dependencies": {
    "express": "^4.21.0",
    "openai": "^4.67.0",
    "redis": "^4.7.0",
    "zod": "^3.23.0"
  }
}

# Initialize .chub/ and detect dependencies in one step
chub init --from-deps

Created .chub/ at ./my-app/.chub
  config.yaml
  pins.yaml
  profiles/base.yaml
  context/architecture.md
  annotations/

Detected 4 dependencies from project files:
  express     javascript (^4.21.0)
  openai      javascript (^4.67.0)
  redis       javascript (^4.7.0)
  zod         javascript (^3.23.0)

Run chub detect --pin to auto-pin detected docs.

2. Auto-pin matching docs

chub detect --pin

Detected 4 dependencies with 3 available docs:

  express  javascript (^4.21.0)  →  express/express  [pinnable]
  openai   javascript (^4.67.0)  →  openai/package   [pinnable]
  redis    javascript (^4.7.0)   →  redis/package    [pinnable]
  ✗ zod

Pinned 3 docs.

Chub matched 3 of 4 deps to curated docs and pinned them. The team now shares the same doc versions.

3. Fetch docs for the AI agent

# Search for relevant patterns
chub search "openai streaming"

5 results for "openai streaming":
  azure/openai      Azure OpenAI JavaScript SDK guide...
  openai/package    openai package guide for Python...
  kafka/streaming   KafkaJS - Apache Kafka client...

# Fetch the pinned OpenAI doc
chub get openai/chat --lang javascript | head -20

# OpenAI API Coding Guidelines (JavaScript/TypeScript)

You are an OpenAI API coding expert. Help me with writing code
using the OpenAI API calling the official libraries and SDKs.

## Golden Rule: Use the Correct and Current SDK

Always use the official OpenAI Node.js SDK.
- NPM Package: `openai`
- Installation: npm install openai

4. Add team knowledge

# Annotate docs with team-specific guidance
chub annotate openai/chat "Always use streaming for chat. Set max_tokens=4096."

Now when any team member or AI agent fetches this doc, they see:

---
[Agent note — 2025-03-21]
Always use streaming for chat. Set max_tokens=4096.

5. Add project context

Create custom docs that the AI agent sees alongside public docs:

cat .chub/context/api-conventions.md

# API Conventions

## Authentication
- All endpoints require Bearer token in Authorization header
- Tokens are JWTs signed with RS256

## Error Responses
- Always return { error: string, code: string } format
- Use HTTP status codes correctly

## OpenAI Integration
- Use streaming for all chat completions
- Set temperature=0.7 for creative tasks, 0 for deterministic

6. Connect to Claude Code

Add to .mcp.json in your project root:

{
  "mcpServers": {
    "chub": {
      "command": "chub",
      "args": ["mcp"]
    }
  }
}

Claude Code now has access to 7 MCP tools:

Tool	What it does
chub_search	Search 1,553+ curated docs
chub_get	Fetch a doc by ID with annotations
chub_list	Browse available docs
chub_context	Get pinned docs, annotations, and project context
chub_pins	Manage pinned doc versions
chub_annotate	Read/write agent notes
chub_feedback	Rate doc quality

When Claude asks "how do I set up Stripe webhooks?", it can search, find the right doc, read your team's annotations, and follow your project conventions — all automatically.

---

Python FastAPI + Stripe project

1. Auto-detect from requirements.txt

cat requirements.txt

fastapi==0.115.0
stripe==10.0.0
uvicorn==0.30.0
pydantic==2.9.0

chub init --from-deps

Detected 4 dependencies from project files:
  fastapi   python (0.115.0)
  stripe    python (10.0.0)
  uvicorn   python (0.30.0)
  pydantic  python (2.9.0)

chub detect --pin

Pinned 4 docs:
  fastapi/package    python
  stripe/payments    python
  uvicorn/package    python
  pydantic/settings  python

2. Create a backend profile

# .chub/profiles/backend.yaml
name: backend
description: Backend API development
includes:
  - fastapi/package
  - stripe/payments
  - pydantic/settings
rules:
  - Use Pydantic v2 model_validator, not v1 validator
  - All Stripe webhook handlers must verify signatures
  - Use async def for all FastAPI endpoints

3. Add Stripe-specific annotations

chub annotate stripe/payments "Use idempotency keys for all POST requests. Webhook endpoint must use raw body parsing — do not use JSONResponse middleware before it."

4. Fetch all pinned docs at once

chub get --pinned

This fetches all 4 pinned docs in one command — ready to paste into a prompt or serve via MCP.

---

Rust project with Cargo.toml

Chub detects dependencies from Cargo.toml (including [workspace.dependencies]):

chub detect

Detected 8 dependencies with 5 available docs:

  tokio      rust (1.40)   →  tokio/runtime     [pinnable]
  serde      rust (1.0)    →  serde/package     [pinnable]
  reqwest    rust (0.12)   →  reqwest/package   [pinnable]
  sqlx       rust (0.8)    →  sqlx/package      [pinnable]
  clap       rust (4.5)    →  clap/package      [pinnable]
  ✗ thiserror
  ✗ anyhow
  ✗ tracing

chub detect --pin    # pin all 5
chub get --pinned    # fetch all at once

---

Supported dependency files

Chub auto-detects from 9 file types:

File	Language	Example
package.json	JavaScript/TypeScript	"express": "^4.21.0"
requirements.txt	Python	fastapi==0.115.0
pyproject.toml	Python	[project.dependencies]
Cargo.toml	Rust	[dependencies] and [workspace.dependencies]
go.mod	Go	require github.com/gin-gonic/gin v1.10
Gemfile	Ruby	gem 'rails', '~> 7.0'
pom.xml	Java	<dependency>
build.gradle	Java/Kotlin	implementation 'org.springframework...'
composer.json	PHP	"require": { "laravel/framework": "^11.0" }