Anthropic's own research shows that structure, not a better model, is what separates a frustrating AI interaction from a remarkable one. flompt turns that insight into a tool: decompose any prompt into 11 structured blocks, edit visually, recompile as Claude-optimized XML.
No sign-up required.
Browser Extension
Build prompts without
leaving the chat.
One install. A ✦ Enhance button appears directly in ChatGPT, Claude and Gemini. Click it to open a structured prompt builder right beside your conversation.
-
Enhance button in every toolbar Injected natively alongside the existing tools. Blends in without breaking the UI.
-
Bidirectional sync Your chat input flows into flompt automatically. Edit your blocks and inject the result back instantly.
-
Resizable sidebar Drag the left edge to adjust the panel width. Stays out of your way when you don't need it.
ChatGPT Claude Gemini
Claude Code integration
Use flompt inside your agent.
flompt ships a native MCP server (Model Context Protocol). Add it once to Claude Code and decompose_prompt, compile_prompt and list_block_types become tools your agent can call directly.
CLI
claude mcp add --transport http --scope user \ flompt https://flompt.dev/mcp/
or ~/.claude.json
{ "mcpServers": { "flompt": { "type": "http", "url": "https://flompt.dev/mcp/" } } }
No install, no account. Hosted on flompt.dev.
Your prompt
You are a helpful assistant. Write me a blog post about AI in healthcare. Make it engaging and professional. Around 800 words. Include statistics. Write in English.
flompt blocks
Role Healthcare tech writer
Objective Blog post · AI in healthcare · 800w
Constraints Professional · Engaging · Data-backed
Output Format Intro → 3 sections → Conclusion
Language English
Assembled prompt Claude-optimized
<prompt> <role> Healthcare tech writer </role> <objective> 800-word blog post on AI in healthcare </objective> <constraints> Professional · 3+ stats </constraints> <examples> <example> <user_input>AI in finance</user_input> <ideal_response>...</ideal_response> </example> </examples> <format_instructions> Use markdown. No preamble. </format_instructions> </prompt>
Prompt anatomy
11 blocks for any prompt.
Every prompt is built from the same ingredients. flompt makes each one explicit, editable, and Claude-optimized.
Document
XML grounding · <document>
Audience
Who reads the output?
Context
Background information
Objective
What to accomplish
Goal
End goal · success criteria
Input
Source material provided
Constraints
Limits and rules
Examples
Structured few-shot pairs
Chain of Thought
Step-by-step reasoning
Output Format
Expected structure
Claude best practices, built-in
Not just decompose. Optimize.
flompt applies Anthropic's official prompting guidelines automatically when you assemble. Every block knows its place.
Document grounding
The Document block wraps your content in Anthropic's <documents><document index="N"> XML format, the structure Claude is trained to parse for highest accuracy.
Structured few-shot examples
The Examples block parses your Input / Output pairs and generates <examples><user_input><ideal_response> XML, the exact format Anthropic recommends for consistent results.
Optimal block ordering
Anthropic's research shows that block order matters. flompt always places documents first, then persona, task, constraints, examples and reasoning. Automatically, without you lifting a finger.
Format Control block
A dedicated block for Claude-specific style directives: be concise, use markdown, no preamble, respond in bullet points. Assembled as <format_instructions> so Claude picks it up reliably.
Block types
12
role · audience · context · objective · goal · input · constraints · examples · chain of thought · output format · document · language
Before
1 blob
↓
After
structured
Semantically decomposed, editable, recompilable
Platforms
3
ChatGPT · Claude · Gemini — inject optimized prompts directly
MIT licensed, free forever
Self-hostable · Docker ready
English & French interface
Browser extension: inject into ChatGPT · Claude · Gemini
Claude Code MCP: use flompt tools natively in your agent