Skip to content

Welcome to Fellow

  • 10x Faster AI Coding


    Stop explaining context. Fellow automatically enriches every coding request with your architecture, patterns, and constraints. Works in CLI and VS Code!

    Quick Start

  • Architectural Guardrails


    AI that respects YOUR constraints. Every suggestion validated against your security, performance, and architectural rules.

    Learn About Guardrails

  • 70% Faster Onboarding


    New developers productive in days, not months. AI assistance that knows your codebase from day 1.

    Onboarding Guide

  • Always Current


    Incremental updates in 10-20 seconds. Knowledge base stays perfectly in sync with your codebase - automatically.

    Incremental Updates

What is Fellow?

Fellow transforms Claude Code from a generic AI assistant into an expert on YOUR specific codebase.

It extracts semantic knowledge from your code - entities, workflows, patterns, and constraints - then automatically enriches every coding request with relevant context. The result: architecturally-correct AI suggestions in seconds, not minutes.

graph LR
    A[Your Codebase] --> B[Fellow Extraction]
    B --> C[Knowledge Base]
    C --> D[Automatic Enrichment]
    D --> E[Claude Code]
    E --> F[Architecturally-Correct Code]

    style B fill:#4051b5
    style C fill:#4051b5
    style D fill:#4051b5
    style F fill:#2e7d32

The Problem Fellow Solves

Without Fellow

You: "Add authentication to this endpoint"

Claude Code: Suggests generic auth code that violates your patterns

You: "No, we use JWT with RS256, store tokens in Redis, check the is_active flag..."

Claude Code: Tries again, still misses your validation rules

Result: 30 minutes wasted explaining context, still get code that needs review

With Fellow

You: "Add authentication to this endpoint"

Claude Code (enriched by Fellow):

📋 Context from Knowledge Base

**Relevant Entities:**
- YOUR AuthService (JWT with RS256)
- YOUR UserModel (with is_active flag)
- YOUR RedisCache (token storage)

**Architectural Guardrails:**
- [Security] Validate JWT signature with RS256
- [Architecture] Use AuthService, not direct DB access
- [Validation] Check user.is_active before auth

Here's the implementation following YOUR patterns...

Result: Architecturally-correct code in 30 seconds

Key Features

Semantic Knowledge Extraction

Extracts three types of knowledge from your codebase:

  • Factual: Entities, models, data structures, relationships
  • Procedural: Workflows, execution flows, integration patterns
  • Conceptual: Architecture style, design patterns, constraints

Automatic Context Enrichment

Every coding request automatically enriched with:

  • ✅ Relevant entities from YOUR codebase
  • ✅ YOUR actual workflows and patterns
  • ✅ YOUR architectural constraints
  • ✅ YOUR design decisions

New: Auto-builds knowledge base if missing - no setup required!

Incremental Updates

  • 10-20 second updates after code changes (vs 3-5 minutes full extraction)
  • Automatic change detection via git diff, file hashes, or modification times
  • Smart merge logic preserves unchanged knowledge
  • 10-20x faster than full rebuild

Hook-Based Automation

  • Automatic enrichment - no manual commands needed
  • Transparent interception of coding requests
  • Pass-through for non-coding questions
  • Enable/disable with /toggle-hooks

Comprehensive Logging

  • JSON and human-readable log formats
  • Tracks every enrichment event
  • Debug detection accuracy
  • Analyze usage patterns

Quick Example

# 1. Build knowledge base (one time setup)
cd /path/to/your-project
/build-kb

# 2. Start coding naturally - Fellow enriches automatically!
# Just type:
"Add rate limiting to the authentication endpoint"

# 3. Update KB after code changes (fast incremental)
/build-kb --update

Why Teams Love Fellow

  • Developers


    "I save 2-3 hours per week. Claude Code finally understands our architecture."

  • Engineering Managers


    "Onboarding time dropped from 12 weeks to 3 weeks. ROI is 27x."

  • Tech Leads


    "Code review time cut by 50%. AI respects our patterns automatically."

  • CTOs


    "We scaled from 5 to 50 engineers without architectural decay."

Who Uses Fellow?

Fellow is trusted by teams at:

  • 🚀 Fast-growing startups - Maintain architecture during rapid growth
  • 🏢 Enterprises - Safe AI for legacy systems and mission-critical code
  • 🔓 Open source projects - Consistent contributions from distributed teams
  • 🏥 Regulated industries - Compliance enforcement in AI-generated code

Get Started in 5 Minutes

# Add as local marketplace
claude plugin marketplace add https://github.com/jingnanzhou/fellow.git

# Install from local marketplace
claude plugin  install fellow@local_marketplace

# Verify installation
claude plugin list
# ✓ fellow@local_marketplace (v2.1.0)
# Navigate to your project
cd /path/to/your-project

# Extract knowledge (3-5 minutes first time)
/build-kb
# Just type your coding requests naturally!
"Add validation to the user registration"

# Fellow enriches automatically with:
# - Your entities
# - Your workflows
# - Your constraints
# After making code changes
/build-kb --update

# Takes 10-20 seconds (incremental)
# Always keeps AI context current

Next Steps

  • Quick Start


    Get Fellow running in 5 minutes with step-by-step instructions.

    Quick Start Guide

  • User Guide


    Learn all commands, features, and configuration options.

    User Guide

  • Use Cases


    See how teams use Fellow for onboarding, legacy code, and more.

    Explore Use Cases

  • Reference


    Quick reference, cheat sheet, and detailed configuration docs.

    Cheat Sheet

Community & Support


Fellow - Architectural Guardrails for Claude Code
Automatically enrich every request to prevent drift and enforce consistency