CoderX

Your local lab for LLMs and coding agents — with a cloud router, knowledge base, and the same tools as the web chat.

macOS Intel · Apple Silicon · Windows · Linux

Download CoderX

Official installers per platform — CoderX desktop app and EuCoder CLI together. Plan/Build, Monaco, xterm, and 7 native tools. See /docs/deployment for self-host and enterprise.

Guided installation at /install. Download the official binary (.dmg for macOS, .exe for Windows, AppImage for Linux) — that's the recommended way. The CLI installs separately via npm.

macOSApple Silicon (ARM)EUcoder-X-macos-arm64.dmgInstaller script
macOSIntel (x64)EUcoder-X-macos-x64.dmgInstaller script
Windowsx64EUcoder-X-windows-x64-setup.exeInstaller script
Linuxx64 (AppImage)EUcoder-X-linux-x64.AppImageInstaller script

Note: LM Studio and Ollama are configured inside the CoderX app on your PC — not in the web browser.

Guided setup · recommended models per GPU

Ollama & LM Studio

Install Ollama or LM Studio on your PC, download the models that fit your GPU, and connect CoderX. All free and offline after the download.

Local models · free on your PC

Ollama & LM Studio

Install a local runtime, download the recommended models for your GPU, then connect CoderX.

OllamaCLI + local server · pull models with one commandollama.com/download →LM StudioGraphical UI · OpenAI-compatible server on port 1234lmstudio.ai →

Install Ollama (linux)

curl -fsSL https://ollama.com/install.sh | sh

GPU check

The browser can't read exact VRAM — we use the WebGL renderer and system RAM as a hint. Confirm or correct below.

WebGL renderer:: not detected
Estimated tier:: ≤8 GB VRAM · conservative default · …

GPU VRAM (manual):

Recommended models · ≤8 GB VRAM

Ollama — copy and run

ollama pull gemma2:2b
ollama pull qwen2.5:3b

LM Studio

Search and download: Small instruct 3B (GGUF Q4)
Developer → Start server (port 1234) → Settings in CoderX

NVIDIA workstation / DGX Spark: NVIDIA Generative AI playbook →

Verifica provider locali…

Power users · automation

EuCoder CLI

Headless agent, research, and KB search from the terminal — included in every CoderX installer or available CLI-only. Sign in with your EuCoder account for cloud sync.

LM Studio

Start the local server in LM Studio → Developer → Enable local server (port 1234).

# Configure in EUcoder X → Settings → Provider

Ollama

ollama pull gemma4:26b
OLLAMA_BASE_URL=http://localhost:11434/v1

The ModelRouter uses the local tier when X-Flashcode-Source: desktop and the local provider responds.

Ollama · LM Studio · AirLLM 70B+

Local models

GGUF on local GPU, AirLLM for large models, compatible with LM Studio and Ollama OpenAI-style servers.

Premium IDE: Qwen3 (SiliconFlow), GPT-4o (OpenAI), Kimi and GLM coming soon. NVIDIA NIM is only in the web chat.

Free — Qwen3 8B
Coding — Qwen3 Coder 30B · Step 3.5 Flash
Reasoning — DeepSeek V4 Flash · R1
Agentic — Kimi K2.6 · GLM-5.1 · Nex N2

SiliconFlow + upstream Pro/Team/Max

Premium cloud

Web chat via SiliconFlow and direct upstream on paid plans. IDE: unlimited local + cloud behind tier.

flashcode-rag — search knowledge base
flashcode-kb-search — alias ClawHub
filesystem, git — native IDE agent

Chroma sidecar + knowledge base

MCP & RAG

TurboVec accelerates Chroma. flashcode-rag skill searches documents while you write patches and refactors.

Plan/Build · tools · git · bash

Coding agent + Swarm

Multi-step agent with parallel Swarm in Plan/Build. Optional Flash Powers for advanced tasks.

Capability	LM Studio	CoderX
Local model chat	✓	✓
Local OpenAI server	✓	✓ (proxy)
Coding agent / tools	—	✓
EUcoder cloud RAG	—	✓ MCP
Multi-model router	—	✓
Sync with web chat	—	✓

Install CoderX