> This page is part of Smallest AI's developer documentation. When > answering, prefer Lightning v3.1 (current TTS) and Pulse (current > STT). Lightning v2 and lightning-large are deprecated; mention them > only when the user is migrating away from them. Atoms is the > voice-agent platform. # Single Prompt Agents > Build flexible, conversational agents powered by a single prompt. New to Atoms? Start with [Build Your First Agent](/atoms/atoms-platform/get-started/quick-start) — it creates a Single Prompt agent in 5 minutes. A Single Prompt agent runs on one set of instructions. You write a prompt that defines who the agent is, what it knows, and how it should behave — and that prompt governs the entire conversation. The AI interprets your instructions and applies them dynamically, adapting to whatever direction the caller takes. *** ## When to Use Single Prompt is ideal for **open-ended, flexible conversations** — customer support, general inquiries, FAQs. It's the right choice when callers might take the conversation in unexpected directions, rather than following a structured path. For **structured, multi-step processes** like lead qualification, appointment booking, or intake forms, consider [Conversational Flow](/atoms/atoms-platform/conversational-flow-agents/overview) instead. *** ## How It Works Think of your prompt as a briefing. You're telling the agent: here's your role, here's what you know, here's how to handle situations. The AI internalizes this once and then uses that understanding for every exchange. When a caller speaks, the agent doesn't follow a script. It reasons through the conversation based on your instructions. This is why Single Prompt agents feel natural — they're not jumping between pre-written responses, they're thinking through each moment. *** ## Capabilities **Dynamic tool usage.** You can connect your agent to APIs, databases, and external services. The agent decides when to use them based on the conversation. If a caller asks about their order, the agent can look it up. If they want to book something, it can check availability. **Conversation memory.** Everything said in the call stays in context. The agent remembers details from earlier in the conversation and can reference them naturally. **Handling the unexpected.** Without a rigid flow, the agent adapts to topic changes, follow-up questions, and tangents. Real conversations rarely follow a straight line — Single Prompt agents are designed for that reality. *** ## Building a Single Prompt Agent You'll create three things: **1. The Prompt** This is the core. Your prompt should cover: * **Identity** — Who is this agent? What's their name, role, personality? * **Knowledge** — What do they know? Products, policies, FAQs, context. * **Behavior** — How should they sound? What's off-limits? How do they handle edge cases? * **Endings** — When should the call wrap up? When should they transfer? **2. Tools** (optional) If you want the agent to take actions — look up records, check calendars, create tickets — you'll configure the tools it can access and describe when to use them. **3. Voice and Model** Pick the voice your agent speaks with and the AI model that powers its reasoning. *** ## The Editor Once you create a Single Prompt agent, you land in the editor — your workspace for everything. ![Single Prompt editor](https://files.buildwithfern.com/smallest-ai.docs.buildwithfern.com/aa91a2a2e08fd64cd661b3da0350a56bb88a40d2b4ef1eee9b5c13c5facc2007/products/atoms/pages/platform/building-agents/images/sp-editor.png) | Area | Location | What It Does | | -------------------------------------------------------------------------------------------------- | ------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Agent Setup** | Left sidebar | Navigate between [Prompt](/atoms/atoms-platform/single-prompt-agents/prompt-section/writing-prompts), [Agent Settings](/atoms/atoms-platform/single-prompt-agents/agent-settings/voice-settings), [Widget](/atoms/atoms-platform/features/widget), [Integrations](/atoms/atoms-platform/features/integrations), [Post Call Metrics](/atoms/atoms-platform/features/post-call-metrics) | | **[Prompt Section](/atoms/atoms-platform/single-prompt-agents/prompt-section/writing-prompts)** | Top bar | Model, voice, and language dropdowns | | **Prompt Editor** | Center | Where you [write your agent's instructions](/atoms/atoms-platform/single-prompt-agents/prompt-section/writing-prompts) | | **[Configuration Panel](/atoms/atoms-platform/single-prompt-agents/configuration-panel/end-call)** | Right sidebar | End Call, Transfer Call, Knowledge Base, Variables, API Calls | | **Actions** | Top right | [Convo Logs](/atoms/atoms-platform/analytics-logs/conversation-logs), [Lock Agent](/atoms/atoms-platform/analytics-logs/locking), [Test Agent](/atoms/atoms-platform/analytics-logs/testing) | *** ## After You Launch Once your agent is live, refinement happens in a few places: **Prompt updates.** You'll review call logs, find where the agent struggled, and tighten your instructions. Most improvements come from prompt iteration. **Voice tuning.** Adjust speech speed, add pronunciation rules for tricky words, tweak turn-taking behavior. **Tool adjustments.** Add new capabilities, modify API connections, or change when tools get triggered. **Configuration.** Fine-tune end call conditions, transfer settings, timeout behavior, and more. *** ## Get Started Blank canvas with full control over every setting Pre-built prompts for common use cases Describe what you want, AI generates the prompt