Configuration

This guide covers all configuration options in Kuse Cowork.

Settings Overview

Access settings by clicking 'Settings' in the bottom-left corner of the application.

AI Provider Configuration

Provider Selection

Kuse Cowork supports multiple AI providers:

Provider	Type	API Key Required
Anthropic	Official API	Yes
OpenAI	Official API	Yes
Google	Official API	Yes
Minimax	Official API	Yes
DeepSeek	Official API	Yes
Ollama	Local	No
LocalAI	Local	No
OpenRouter	Aggregator	Yes
Together AI	Aggregator	Yes
Groq	Aggregator	Yes
SiliconFlow	Aggregator	Yes

API Key Management

API keys are stored locally in an encrypted SQLite database. Keys are never sent to any server other than the respective AI provider.

~/.kuse-cowork/settings.db

API Key Safety

• Keys are stored locally only
• No telemetry or analytics
• Keys are passed directly to provider APIs

Model Selection

Each provider offers multiple models:

Anthropic:

Model	Description
claude-opus-4-5-20251101	Most capable
claude-sonnet-4-5-20250929	Balanced (recommended)

OpenAI:

Model	Description
gpt-5	Latest flagship
gpt-5-mini	Fast and efficient
gpt-4o	Multimodal

Local (Ollama):

Model	Description
llama3.3:latest	Meta's latest open source
qwen2.5:latest	Good for Chinese
codellama:latest	Code-specialized
deepseek-r1:latest	Strong reasoning

Custom Endpoints

For self-hosted or enterprise deployments:

1. Select "Custom Service" as provider
2. Enter your base URL (e.g., http://localhost:8000)
3. Configure authentication if needed
4. Enter your model ID

Application Settings

Max Tokens

Controls the maximum length of AI responses:

• Default: 4096 tokens
• Range: 256 - 128000 tokens
• Recommendation: 4096 for chat, 8192+ for code generation

Temperature

Controls response randomness:

• 0.0: Deterministic, focused
• 0.7: Balanced (default)
• 1.0: Creative, varied

Reasoning Models

GPT-5, o1, and o3 models don't support custom temperature.

Project Configuration

Project Folder

Set the working directory for tasks:

1. Click "Select Folder" in task creation
2. Browse to your project root
3. The agent will have access to all files in this directory

Docker Mounting

The project folder is mounted at /workspace in Docker containers.

Skills Directory

Custom skills are stored in:

~/.kuse-cowork/skills/

See Skills Guide for creating custom skills.

Docker Configuration

Container Settings

Kuse Cowork uses Docker for safe command execution:

Setting	Value
Default Image	python:3.11-alpine
Available Images	ubuntu:latest, node:20, rust:alpine
Workspace Mount	/workspace
Skills Mount	/skills (read-only)

Verifying Docker

Check Docker status in Settings. If disconnected:

1. Ensure Docker Desktop is running
2. Check Docker socket permissions
3. Restart Kuse Cowork

MCP Configuration

Adding MCP Servers

1. Go to Settings → MCP
2. Click "Add Server"
3. Enter server configuration:

{
  "name": "my-mcp-server",
  "url": "http://localhost:3000",
  "auth": {
    "type": "bearer",
    "token": "your-token"
  }
}

OAuth MCP Servers

For OAuth-authenticated MCP servers:

1. Click "Connect with OAuth"
2. Complete authentication in browser
3. Token is stored locally

Data Storage

All data is stored locally:

Data	Location
Settings	~/.kuse-cowork/settings.db
Conversations	~/.kuse-cowork/settings.db
Tasks	~/.kuse-cowork/settings.db
Skills	~/.kuse-cowork/skills/

Backup

To backup your data:

cp -r ~/.kuse-cowork ~/.kuse-cowork-backup

Reset

To reset all settings:

rm -rf ~/.kuse-cowork

Environment Variables

For advanced configuration, environment variables can be used:

Variable	Description
KUSE_DATA_DIR	Override data directory
KUSE_LOG_LEVEL	Set log level (debug, info, warn, error)

Configuration File

Settings can also be configured via JSON:

// ~/.kuse-cowork/config.json
{
  "defaultProvider": "anthropic",
  "defaultModel": "claude-sonnet-4-5-20250929",
  "maxTokens": 4096,
  "temperature": 0.7,
  "dockerEnabled": true
}

Next Steps

• AI Providers - Detailed provider documentation
• Skills - Create custom skills
• MCP Protocol - Connect external tools