Skip to main content

Agent Settings Guide

This guide covers everything you need to configure AI agents in Wegent through the Web interface. An Agent combines Bots, collaboration modes, and AI models to accomplish tasks.

πŸ“‹ Table of Contents​

🎯 Overview​

What is an Agent?​

An Agent is your AI assistant in Wegent. It can:

  • Execute development tasks
  • Answer questions
  • Review code
  • Generate documentation
  • And much more...

Architecture​

Agent = Bot(s) + Collaboration Mode

Bot = Executor + Model + Prompt + Tools
ComponentDescriptionExample
AgentUser-facing AI assistant"Code Assistant"
BotBuilding block of an Agent"Frontend Developer Bot"
ExecutorRuntime environmentClaudeCode, Dify, Chat
ModelAI brainClaude Sonnet 4, GPT-4o
PromptPersonality & expertiseSystem instructions
ToolsExternal capabilitiesGitHub MCP, Skills

βš™οΈ Accessing Agent Settings​

Via Web Interface​

  1. Log in to Wegent
  2. Click Settings in the sidebar
  3. You'll see these tabs:
TabDescription
TeamManage your agents
BotManage individual bots
ModelsConfigure AI models
ShellsCustom executors
SkillsChat and Claude Code skills
IntegrationsGit tokens

Quick Access​

  • From the chat interface, click the gear icon next to the agent selector
  • Or click Manage in the agent dropdown

πŸš€ Creating an Agent​

Step 1: Open Creation Dialog​

  1. Go to Settings β†’ Team tab
  2. Click New Team button

Step 2: Basic Information​

FieldRequiredDescription
NameYesAgent name (e.g., "Code Assistant")
DescriptionNoWhat this agent does
IconNoVisual identifier
Bind ModeYesWhere agent appears (Chat/Code)
Requires RepositoryNoNeed code repo to work?

Bind Mode Options​

ModeDescription
ChatAppears in Chat mode for conversations
CodeAppears in Code mode for development
BothAvailable in both modes

Step 3: Select Collaboration Mode​

Choose how your agent works:

ModeBotsBest For
Solo1Simple tasks, beginners
Pipeline2+Sequential workflows
Coordinate2+Parallel analysis

The web UI currently offers Solo, Pipeline, and Coordinate when creating or editing agents.

Recommendation: Start with Solo mode for simplicity.

Step 4: Configure Bot​

For Solo mode, configure the bot directly:

  1. Name: Bot identifier
  2. Executor: Select runtime (ClaudeCode, Chat, Dify)
  3. Bind Model: Select AI model
  4. Prompt: Define personality and expertise
  5. MCP Config: Add external tools (optional)
  6. Skills: Add capabilities (optional)

For other modes, select existing bots or create new ones.

Step 5: Save​

Click Save to create your agent. It will appear in your agent list.

πŸ€– Configuring Bots​

Bot Components​

ComponentRequiredDescription
NameYesUnique identifier
ExecutorYesRuntime environment
ModelNoAI model (can inherit)
PromptNoSystem instructions
MCP ConfigNoExternal tools
SkillsNoChat and Claude Code skills

Executor Types​

ExecutorDescriptionUse Case
ClaudeCodeClaude Code SDK in DockerCode development
ChatDirect LLM APISimple conversations
DifyExternal Dify APIDify workflows

Writing Effective Prompts​

Structure:

You are a [role], skilled in [expertise].

Your responsibilities:
- [Task 1]
- [Task 2]

Working principles:
- [Guideline 1]
- [Guideline 2]

Example:

You are a senior frontend engineer, skilled in:
- React 18+ and TypeScript
- Tailwind CSS
- Performance optimization

Your responsibilities:
- Develop high-quality UI components
- Write clean, maintainable code
- Follow React best practices

Working principles:
- Use functional components and Hooks
- Write TypeScript types for all components
- Focus on user experience

Managing Bots​

  1. Go to Settings β†’ Bot tab
  2. View all your bots
  3. Click Edit to modify, Delete to remove
  4. Click New Bot to create standalone bots

🀝 Collaboration Modes​

Solo Mode​

Single bot handles everything.

User β†’ Bot β†’ Result

Best for: Simple tasks, getting started

Pipeline Mode​

Bots execute sequentially.

Bot A β†’ Bot B β†’ Bot C β†’ Result

Best for: Development β†’ Review β†’ Testing workflows

Configuration:

  1. Select Leader bot (first in pipeline)
  2. Add Member bots in order
  3. Optionally enable Require Confirmation between stages

Coordinate Mode​

Leader coordinates parallel work.

        β†’ Bot A (parallel)
Leader β†’ Bot B (parallel) β†’ Aggregate
        β†’ Bot C (parallel)

Best for: Multi-perspective analysis

🧠 Model Configuration​

Binding Models to Bots​

Method 1: Dropdown Selection (Recommended)

  1. In Bot configuration, find Bind Model
  2. Keep "Advanced Mode" OFF
  3. Select from available models

Method 2: Advanced Mode

  1. Toggle Advanced Mode ON
  2. Select Model Protocol (OpenAI/Claude/Gemini)
  3. Enter JSON configuration

Creating Custom Models​

  1. Go to Settings β†’ Models tab
  2. Click Create Model
  3. Configure:
    • Model Type (LLM/Embedding/Rerank)
    • Protocol (OpenAI/Anthropic/Gemini)
    • API Key
    • Model ID

See Model Configuration Guide for details.

Per-Task Model Override​

When sending a task:

  1. Click the Model selector in chat input
  2. Choose a different model
  3. Enable Force Override if needed

πŸ”§ MCP Tools & Skills​

MCP (Model Context Protocol) Tools​

MCP tools extend bot capabilities with external services.

Adding MCP Tools:

  1. In Bot configuration, find MCP Config
  2. Click Import MCP
  3. Paste configuration JSON
  4. Choose Replace or Append

Example - GitHub MCP:

{
  "github": {
    "command": "docker",
    "args": [
      "run", "-i", "--rm",
      "-e", "GITHUB_PERSONAL_ACCESS_TOKEN",
      "ghcr.io/github/github-mcp-server"
    ],
    "env": {
      "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_xxx..."
    }
  }
}

Skills​

Skills add reusable capabilities to Chat and Claude Code bots.

Adding Skills:

  1. In Bot configuration, find Skills section
  2. Select skills from dropdown
  3. Check the preload option next to a selected skill when it should be available from task start
  4. Click Manage Skills to upload new ones

Preloading Skills:

  • Both the simple configuration UI and the full configuration UI support preload skills
  • Chat and ClaudeCode executors support preload skills
  • For ClaudeCode, preloaded skills are deployed into the Claude Code skills directory when the task starts
  • Preload only core skills that the bot should prioritize; leave other skills on-demand

Managing Skills:

  1. Go to Settings β†’ Skills tab
  2. Upload ZIP packages containing SKILL.md
  3. Or import from Git repositories

✨ Best Practices​

1. Start Simple​

  • Begin with Solo mode
  • Use one bot with clear responsibilities
  • Add complexity as needed

2. Choose Right Executor​

TaskExecutor
Code developmentClaudeCode
Simple Q&AChat
Dify workflowsDify

3. Write Clear Prompts​

  • Be specific about the role
  • List concrete responsibilities
  • Include working guidelines

4. Optimize Costs​

ComplexityModel
SimpleClaude Haiku / GPT-3.5
MediumClaude Sonnet / GPT-4o
ComplexClaude Opus / GPT-4

5. Use Descriptive Names​

βœ… "Frontend React Developer"
βœ… "Code Review Expert"
❌ "Bot 1"
❌ "Test"

❓ FAQ​

Q: What's the difference between Agent and Bot?​

  • Agent: User-facing AI assistant (what you interact with)
  • Bot: Building block of an Agent

An Agent can have one Bot (Solo) or multiple Bots (other modes).

Q: Which collaboration mode should I use?​

ScenarioMode
Simple tasksSolo
Sequential workflowPipeline
Parallel analysisCoordinate

Q: Can I edit an Agent after creation?​

Yes! Click Edit on any agent to modify it.

Q: Why doesn't my Agent appear in Chat/Code?​

Check Bind Mode setting - enable Chat and/or Code as needed.

Q: How do I share an Agent?​

  1. Click Share on the agent
  2. Copy the share link
  3. Others can add it to their list

Q: What's "Requires Repository"?​

  • Enabled: Must select a code repo before starting
  • Disabled: Can work without a repo

Enable for development agents, disable for chat agents.

πŸ”— Related Resources​

Configure your AI agents and start automating! πŸš€