User Guide

Comprehensive guides for using Traigent SDK features and capabilities.

Contents

• Agent Optimization - Optimize AI agents and multi-step workflows
• Execution Modes - The three modes (local, cloud, hybrid_api): where your agent runs, who picks configurations, what leaves the machine
• Choosing Models - Select the right algorithm / offline routing model for your use case
• Configuration Spaces - Define tunable variables, parameter ranges, and presets
• Tuned Variables - Auto-detect tunable parameters and use the callable discovery API
• Injection Modes - Different ways to integrate Traigent with your code
• Interactive Optimization - Real-time optimization workflows and monitoring
• Evaluation Guide - Measure and improve model performance
• Smart Optimization - Use backend-routed smart optimizer strategies

Learning Path

New Users: Start with Injection Modes to understand integration options

Basic Optimization: Learn Choosing Models for your use case

Tuned Variables: Use Tuned Variables to auto-detect parameters and discover callables

Advanced Users: Explore Agent Optimization, Interactive Optimization, and Smart Optimization

Evaluation: Use Evaluation Guide to measure success

CLI & Observability Tools

• Onboarding: traigent onboard, browser traigent auth device-login, traigent quickstart, and traigent first-prompt --agent claude|cursor|codex|copilot.
• Local MCP: traigent mcp serve provides the stdio MCP server for coding agents.
• Observability: add_agent_span() adds user-defined agent workflow spans; see telemetry.

See CHANGELOG.md for what's new in each release.