Skip to content

Agent Tools

Tools for invoking custom AI agents.

Overview

Nexus-Dev supports custom agents—specialized AI personas defined in YAML files. Each agent becomes an MCP tool that can be invoked by your IDE's AI assistant.

flowchart LR
    IDE["🤖 Your IDE"]
    NX["📡 Nexus-Dev"]
    Agent["👤 Custom Agent"]
    RAG["📚 RAG"]

    IDE -->|"ask_code_reviewer(...)"| NX
    NX --> Agent
    Agent --> RAG
    Agent -->|response| NX
    NX -->|result| IDE

refresh_agents

Reload custom agents from the agents/ directory.

When to Use

  • After creating a new agent
  • After modifying an agent configuration
  • When the MCP server started without project context

Example

refresh_agents()

Response:

Refreshed agents from /path/to/project/agents/

Registered tools:
- ask_code_reviewer
- ask_security_auditor
- ask_my_custom_agent

list_agents

Show available agents in the current workspace.

Example

list_agents()

Response:

## Available Agents

### code_reviewer
- **Role:** Senior Code Reviewer
- **Description:** Delegate code review tasks to the Code Reviewer agent.

### security_auditor
- **Role:** Security Expert
- **Description:** Identify vulnerabilities and recommend fixes.

ask_* (Dynamic Tools)

Each custom agent gets its own ask_<name> tool.

Parameters

Parameter Type Required Description
task string ✅ Task description for the agent

Example

ask_code_reviewer("Review the authentication module for security issues and best practices")

Response:

The agent processes the task using:

  1. Its configured persona (role, goal, backstory)
  2. RAG search results from the knowledge base
  3. Any configured tools

And returns a structured response based on its expertise.

Creating Agents

From Template

nexus-agent init my_reviewer --from-template code_reviewer

Available Templates

Template Role
code_reviewer Reviews code for bugs and best practices
doc_writer Creates technical documentation
debug_detective Analyzes errors and proposes fixes
refactor_architect Suggests code restructuring
test_engineer Generates test cases
security_auditor Identifies vulnerabilities
api_designer Reviews REST/GraphQL APIs
performance_optimizer Finds bottlenecks

Agent Configuration

Agents are defined in agents/<name>.yaml:

name: "code_reviewer"
display_name: "Code Reviewer"
description: "Delegate code review tasks to the Code Reviewer agent."

profile:
  role: "Senior Code Reviewer"
  goal: "Identify bugs, security issues, and suggest improvements"
  backstory: "Expert developer with 10+ years of experience."
  tone: "Professional and constructive"

memory:
  enabled: true
  rag_limit: 5
  search_types: ["code", "documentation", "lesson"]

tools: []  # Empty = all tools available

llm_config:
  model_hint: "claude-sonnet-4.5"
  fallback_hints: ["auto"]
  temperature: 0.5
  max_tokens: 4000

Configuration Options

Field Description
profile.role Agent's expertise/persona
profile.goal What the agent tries to achieve
profile.backstory Background for context
profile.tone Response style
memory.enabled Use RAG for context
memory.rag_limit Number of RAG results
memory.search_types What to search
llm_config.model_hint Preferred model

Usage Patterns

Code Review

ask_code_reviewer("Review src/auth/login.py for security issues")

Documentation

ask_doc_writer("Create API documentation for the UserService class")

Debugging

ask_debug_detective("Analyze this error: ConnectionRefusedError on port 5432")

See Also