AI Agents Overview

Friday Dev integrates 9+ AI agents, each optimized for different development tasks.

Available Agents

Agent	Provider	Best For	Cost
Friday Agent	Anthropic	Complex tasks	API usage
Claude	Anthropic	Safe, detailed	API usage
Gemini	Google	Fast, free tier	Free/API
GPT-5 Codex	OpenAI	Advanced reasoning	API usage
GLM 4.7	Zhipu AI	Multilingual	API usage
Qwen Code	Alibaba	Coding focus	API usage
Cursor	Cursor	IDE integration	Subscription
GitHub Copilot	GitHub	Code completion	Subscription
Droid	Multiple	Flexible	Varies

Comparison

┌─────────────────────────────────────────────────────────────────┐
│                     Agent Capabilities                           │
├──────────────┬──────────┬──────────┬──────────┬────────────────┤
│    Agent     │  Speed   │ Quality  │  Safety  │    Cost        │
├──────────────┼──────────┼──────────┼──────────┼────────────────┤
│ Friday       │   ██░░   │  █████   │  █████   │  $$$           │
│ Claude       │   ██░░   │  █████   │  █████   │  $$$           │
│ Gemini       │   ████   │  ████░   │  ████░   │  Free/$        │
│ GPT-5 Codex  │   ███░   │  █████   │  ████░   │  $$$$          │
│ GLM 4.7      │   ████   │  ████░   │  ████░   │  $$            │
│ Qwen         │   ████   │  ████░   │  ███░░   │  $$            │
│ Droid        │   ███░   │  ████░   │  ███░░   │  Varies        │
└──────────────┴──────────┴──────────┴──────────┴────────────────┘

Quick Selection Guide

By Task Type

Task	Recommended Agent
Complex feature implementation	Friday Agent
Bug fixes	Gemini (fast) or Claude (thorough)
Code refactoring	Claude
Documentation	Gemini
Security-sensitive code	Claude
Quick prototypes	Gemini
Multilingual projects	GLM 4.7

By Priority

Priority	Agent
Speed	Gemini > GLM > Qwen
Quality	Friday = Claude > GPT-5
Safety	Claude > Friday > Others
Cost	Gemini (free) > GLM > Others

How Agents Work

1. Task Analysis

Agent reads and understands:

• Task description
• Codebase structure
• Relevant files
• Project patterns

2. Planning

Agent creates a plan:

• Files to modify
• Changes to make
• Order of operations

3. Execution

Agent performs:

• Code writing
• File modifications
• Command execution
• Testing (if available)

4. Output

Agent delivers:

• Code changes (diff)
• Execution logs
• PR (if configured)
• Status update

Autonomy Levels

Level	Description	Risk
workspace-write	Only modify workspace files	Low
skip-permissions-unsafe	Full system access	High

Recommendation: Use workspace-write for most tasks.

API Keys

Most agents require API keys:

# Set API keys
export GEMINI_API_KEY="your-key"
export OPENAI_API_KEY="your-key"
export ANTHROPIC_API_KEY="your-key"

See individual agent pages for setup instructions.

Running Agents

UI

1. Open task
2. Click "Run Agent"
3. Select agent
4. Configure options
5. Start

CLI

# Run specific agent
friday-dev run --task 123 --agent friday

# Run with options
friday-dev run --task 123 --agent gemini --autonomy workspace-write

Monitoring

Live Logs

Watch agent progress:

friday-dev logs --task 123 --follow

Status Dashboard

View all running agents in the dashboard.

Best Practices

1. Start with Gemini for simple tasks (fast, free)
2. Use Friday/Claude for complex features
3. Always review agent output before merging
4. Use workspace-write autonomy for safety
5. Provide clear task descriptions

Troubleshooting

Agent Stuck

• Check API key validity
• Verify network connection
• Try a different agent
• Break task into smaller parts

Poor Output

• Improve task description
• Add context in description
• Try a different agent
• Use subtasks

Rate Limits

• Wait and retry
• Use different API key
• Switch agents
• Self-host for unlimited

Next Steps

• Friday Agent - Learn about the flagship agent
• Task Management - Write better tasks
• Configuration - Set up API keys