Built-in Servers

Everything You Need, Built In

Other tools require you to set up external services. Install Ollama. Configure a vector database. Run a separate embedding server. FrankenCoder includes everything:

fc-inference

Our built-in LLM inference server:

• GGUF model support: Run quantized models efficiently
• GPU acceleration: CUDA, Metal, Vulkan
• CPU fallback: Works without GPU too
• OpenAI-compatible API: Drop-in replacement
• Model management: Download, load, switch models

fc-embed

Local embedding generation for RAG:

• CUDA: NVIDIA GPU acceleration
• WGPU: Cross-platform GPU (AMD, Intel)
• CPU: Works on any machine
• BGE-small: Optimized for code
• Fast: Thousands of embeddings per second

Indexing Server

Continuous background indexing:

• Real-time: Index as you code
• Incremental: Only re-index changes
• File watching: Automatic detection
• Per-workspace: Isolated indexes

Agent Daemon

The orchestration layer for all 13 agents:

• Task distribution: Routes work to right agents
• Context sharing: Agents communicate seamlessly
• Loop management: Autonomous coding cycles
• Queue handling: FIFO for multi-tab work

Zero Configuration

Install FrankenCoder. Launch it. Everything works. No Docker, no config files, no environment variables. Just code.