Agent Settings

General Settings

Configure core behavioral defaults and execution limits.

Auto-apply changes Apply modified codebase files automatically without launching a side-by-side diff preview.

Allow terminal commands without confirm Allow the agent to automatically execute shell commands in your terminal without requiring manual verification.

Skip Ollama when @files mentioned Skip running the local Ollama file selector when explicit @file context references are present.

Hybrid Routing

Enable Hybrid Routing Route simpler pipeline tasks to a local Ollama instance and complex code generation to the main provider.

Complexity Threshold

Limits & Context

Response Timeout (minutes) Conversation Turns Context Max Stored History Sessions

Diagnostics & Auto-Fix

Auto-fix compilation errors Automatically trigger a follow-up code modification loop to fix compilation errors/warnings introduced by agent modifications.

Max Auto-Fix Retries

LLM Providers

Configure API keys, model parameters, and provider backends.

Active LLM Backend Choose which LLM source handles code generation and file selection. File Selection Backend Which LLM backend to use for choosing files relevant to your task. Active Browser Tab AI Provider Select which browser tab AI to bridge to (ChatGPT, Gemini, or Claude). Only relevant for Bridge mode.

Ollama Settings

Ollama HTTP URL Ollama Model

Custom OpenAI Compatible API

Quick Presets Instantly configure parameters for popular local clusters or tools.

Base URL API Key Model Name Provider Label

AgentRouter API

AgentRouter Base URL API Key Model Name

Groq API

Groq Base URL API Key Get a free key at console.groq.com/keys. Model Name

Swarm Settings

Swarm Broker URL Device Display Name

Contribute Compute Resources Allow this machine to process assigned model layers (uses GPU VRAM, or fallback RAM).

Hugging Face Access Token Required for downloading gated or private models (e.g. Llama 3). Custom Swarm Models Enter a comma-separated list of Hugging Face model IDs. Hugging Face Cache Folder Path Custom directory where Hugging Face downloads models. Swarm Worker Python Path Custom Python path or command for the Swarm worker.

Downloaded Models (Local Cache) Models cached on this machine via Hugging Face. Click on any model to copy its ID.

No models found in local cache (or worker not connected)

Download & Cache a Model Enter a Hugging Face repo ID to pre-download/cache it on this machine (e.g. TinyLlama/TinyLlama-1.1B-Chat-v1.0).

Python & Library Setup Guide

1. Recommended Python Versions Python 3.9 to 3.12 are highly recommended for optimal PyTorch and CUDA support. Python 3.13 is supported in PyTorch 2.5.1+. Make sure to install a 64-bit version of Python.

2. Install Required Packages Run this command in your terminal to install the necessary libraries:

pip install torch transformers huggingface_hub psutil

3. Locate Your Python Executables Run these commands in your shell to find the exact paths to your installed Python executables:

Windows (PowerShell)

where.exe python

macOS / Linux (Terminal)

which -a python3

General Settings

LLM Providers

Swarm Groups

--