Generate expert-crafted prompts built specifically for Claude's unique reasoning, analysis, and creative capabilities with system prompts, technique tags, and Claude-specific tips included. Free, instant, no account needed.
Built around Claude's unique strengths — extended reasoning, document analysis, structured XML output, and nuanced instruction-following.
Tell us what you want Claude to do. Add tone, goal, audience, and format to guide the output. Even a brief description works well.
Choose which AI writes your Claude prompts — Grok, Gemini 2.0, Amazon Nova, or NVIDIA Nemotron. Each offers a different creative angle.
Receive up to 6 prompts tailored to Claude's capabilities — with XML structure, system prompts, technique tags, and Claude-specific expert tips.
One-click copy and paste directly into Claude.ai, the Claude API, or any Claude-powered application. No cleanup needed.
Claude thinks differently from other LLMs. These prompts are engineered to unlock what makes Claude exceptional.
Claude is uniquely trained to parse and respect XML tags. Every complex prompt uses structured tags like <context>, <task>, and <output_format> — a technique that doesn't work as well in other models.
Prompts include language that activates Claude's chain-of-thought reasoning — "think step by step," "use a thinking section," and similar techniques that improve accuracy on complex tasks.
Each result includes a tailored system prompt for the Claude API and Claude-powered applications — defining behavior, persona, and rules precisely for your use case.
Claude excels at long-document analysis. Our prompts are structured to maximize Claude's ability to extract, synthesize, and reason about complex documents.
Claude responds to richer persona definitions than most models. Prompts assign specific expertise, communication style, and decision-making frameworks — not just a job title.
Each prompt card includes one expert tip specific to Claude — calibrated to Claude's behavior, training, and the types of requests it handles best.
Hand-crafted, tested Claude prompts across every category. Click any prompt to copy it.
Everything you need to know about getting the best results from Claude.
Generate expert-engineered Claude prompts in seconds. Free, instant, no account required.
Generate Free Prompts NowClaude is Anthropic's AI assistant, built with a distinct architecture and training philosophy. Unlike other large language models, Claude is designed with a Constitutional AI framework — it reasons carefully, handles nuance exceptionally well, and follows complex instructions with high fidelity. These characteristics mean the quality of your Claude prompt has an outsized effect on output quality: a mediocre prompt gives you a mediocre response; a carefully engineered prompt unlocks Claude's remarkable reasoning and analytical depth.
Our Claude prompt generator is purpose-built for Claude's unique capabilities — not adapted from generic prompt templates. Every generated prompt uses techniques that specifically amplify what Claude does best.
This is the single most Claude-specific technique in prompt engineering. Claude is trained to parse and respect XML tags, allowing you to clearly separate different parts of your prompt. Wrap background context in <context> tags, your task in <task> tags, examples in <examples> tags, and specify output in <output_format> tags. This prevents Claude from confusing instructions with content and produces dramatically more accurate responses on complex, multi-part requests.
Unlike models that work best with terse prompts, Claude is trained on long, complex documents. It handles rich context gracefully. Tell Claude everything relevant: who you are, what the business situation is, what has already been tried, what constraints exist, and who the final output is for. The more relevant context you provide, the less Claude has to infer — and inference introduces error.
Claude's role assignment goes deeper than other models. Instead of "you are a marketing expert," try "You are a B2B SaaS marketing director with 12 years of experience in enterprise software, specializing in demand generation and account-based marketing. You communicate in a strategic, data-informed way and always ground recommendations in specific tactics rather than general principles." This level of specificity genuinely shapes Claude's reasoning approach.
For any task requiring multi-step reasoning, analysis, or planning, add thinking-activation language to your prompt: "Before giving your final response, think through this step by step in a <thinking> section." This forces Claude to reason systematically before producing output, which significantly improves accuracy on complex tasks — equivalent to showing your work in math.
Tell Claude exactly how to structure its response. "Respond as a structured JSON object with fields: title, executive_summary, key_findings (array), and recommended_actions (array)" gets you a perfectly formatted, parseable output. For prose, specify structure: "Organize your response with these sections: Background, Analysis, Recommendations, and Next Steps. Use H2 headers for each section." Claude's instruction-following is precise enough to handle very specific format requirements.
In the Claude API, you can prefill the beginning of Claude's response — essentially starting its answer for it. This is a powerful technique for forcing a specific format or preventing Claude from adding unnecessary preamble. For example, prefilling with "{" forces Claude to output valid JSON. Prefilling with "The three primary risks are:" shapes the response structure from the start.
Claude's training means it naturally considers ethical implications, acknowledges uncertainty, and provides balanced perspectives. For research and analysis tasks, you can leverage this: "Consider multiple perspectives before forming your conclusion" or "Identify potential weaknesses in your own analysis." Claude will genuinely do this, producing more nuanced and reliable analysis than models optimized purely for confident-sounding output.
Claude's 200,000-token context window makes it uniquely capable for document analysis. For best results, structure your document analysis prompts with three components: the document itself (or a specific section), precise extraction instructions ("identify all claims about market size, preserving the exact phrasing used"), and an output format specification. The more explicitly you define what to extract and how to present it, the more accurate and useful Claude's analysis becomes.
The best Claude prompts are designed for iteration, not single-shot completion. Structure your prompt to explicitly invite follow-up: "After your initial response, I will provide specific feedback or additional context. Please ask me one clarifying question if something about my request is ambiguous." This transforms Claude from a response machine into a collaborative reasoning partner — which is its most powerful mode.
Different tasks require different prompt architectures for Claude. Here are the foundations:
Document analysis: Role + document context + specific extraction instructions in XML + output format specification. Example: "You are an expert legal analyst. <document>[document text]</document> <task>Identify all payment obligations, their amounts, due dates, and which party bears each obligation.</task> <output_format>A structured table with columns: Obligation, Amount, Due Date, Responsible Party, Consequence of Non-Payment.</output_format>"
Complex analysis: Role + context + thinking activation + structured output. Example: "You are a senior strategy consultant. <context>[situation description]</context> Think step-by-step through this problem in a <thinking> section, then provide your recommendations structured as: Problem Definition, Root Causes (top 3), Strategic Options (2-3), Recommended Path with rationale, and Implementation Risks."
Code generation: Language + requirements + code style + error handling instructions + output format. Claude produces cleaner code when you specify: "Include inline comments for non-obvious logic, use descriptive variable names, add a docstring explaining inputs and outputs, and handle the most likely error cases with appropriate exceptions."
Creative writing: Genre + protagonist description (including a specific flaw) + narrative constraints + tone + length. For Claude's creative work, adding "Do not default to the most obvious interpretation of this premise — find the unexpected angle" consistently produces more distinctive, memorable output.