> ## Documentation Index > Fetch the complete documentation index at: https://docs.lyzr.ai/llms.txt > Use this file to discover all available pages before exploring further. # Agent Builder (Conversational AI) The **Lyzr Agent Builder** is a sophisticated, natural language interface designed to transition ideas into runnable, enterprise-grade agents in minutes. Positioned on the Lyzr Studio home page, it serves as a high-speed alternative to manual configuration, empowering both developers and business users to ship production-ready agents without starting from scratch. Lyzr Home Page Agent Builder *** ## The Conversational Build Logic The builder uses an intelligent, iterative process to handle the heavy lifting of agent architecture, moving beyond simple chat-bot creation into full autonomous system design. ### 1. Initiating the Build On the Studio Home screen, use the central text area to describe the agent you want to create. * **Natural Language Input:** Simply describe the role and task (e.g., *"Build a Stock market analyst agent"*). * **Architect Integration:** Access advanced templates and organizational blueprints by clicking **Explore Architect**. ### 2. Feature Requirement Gathering Once you submit your prompt, the **Lyzr Agent Builder** (Beta) initiates a specialized chat interface. It analyzes your request and presents a series of follow-up questions or checkboxes to refine the agent's scope. For a specialized agent, the builder might suggest and verify: * **Knowledge Requirements:** Does the agent need a RAG pipeline for specific documents? * **Tool Access:** Should the agent have permissions to fetch real-time data or interact with APIs? * **Reasoning Depth:** Identifying if the task requires high-reasoning models like o1 or Claude 3.5 Sonnet. ### 3. Automated Infrastructure Assembly Once sufficient context is gathered, the system automatically handles the "heavy lifting" of the backend setup: * **Model Selection:** Chooses the optimal LLM provider and model based on the complexity of the task. * **Knowledge Provisioning:** Automatically creates or connects a **Knowledge Base** (RAG) or **Knowledge Graph**. * **Tool Orchestration:** Identifies and connects relevant external tools (e.g., Slack, GitHub, or custom APIs). * **Multi-Agent Setup:** Capable of establishing **Managerial Orchestration** to delegate sub-tasks to specialized worker agents. *** ## Core Capabilities & Customization ### Modular Building Blocks While the builder automates the initial setup, you retain full control to refine the output through the Studio dashboard: * **Manual Overrides:** At any point, you can transition to the granular **Agent Builder** screen for manual tweaking of prompts and features. * **Skill Attachment:** Enhance agents with modular **Skills**—reusable blocks of logic or tool integrations that make agents production-ready. * **Memory Management:** Toggle between **Lyzr Cognis** (cross-session memory) and **Lyzr Memory** (adjustable short/long-term summarization). Agent Management Interface ### Multi-Channel Deployment Every agent built via the conversational path is instantly ready for enterprise use: * **Web/Mobile Mini-Apps:** Package your agent as a shareable app for internal teams or customers. * **API-First Integration:** Every agent auto-generates an **OpenAPI 3.1 endpoint** and gRPC stubs, ready for microservices or protocols like **MCP**. *** ## Integrated Governance & Testing Lyzr ensures that agents built via natural language remain safe through native integration with the **Safe & Responsible Module**. ### Simulation & Agent Hardening * **Automated Scenarios:** The system generates realistic test cases and "personas" to stress-test your agent's logic. * **Metric Scoring:** Responses are evaluated against metrics like **Task Completion, Hallucination, and Faithfulness**. * **Hardening:** Lyzr analyzes failures and recommends optimal configuration updates to improve accuracy before production rollout. ### Responsible AI Guardrails Activate the **Safe/Responsible-AI lattice** with a single click to ensure compliance and ethical operation: * **PII Redaction:** Automatically masks sensitive data like Credit Card numbers, SSNs, and personal emails. * **Prompt Injection Protection:** Detects and blocks malicious attempts to override agent instructions. * **Hallucination Manager:** Employs a **Reflection Mechanism** where the agent critically assesses its own response against instructions before delivery. *** ## Next Steps Once the Builder completes the setup, you can perform the following actions to finalize your deployment: 1. **Review in Studio:** Fine-tune the generated Agent Role, Goal, and Instructions. 2. **Test in Simulation:** Run a batch of simulations to see performance against "edge-case" user inputs. 3. **Deploy:** Connect your agent to your preferred interface (Web, Slack, or Twilio) using the auto-generated API keys.