Multi-LLM Prompt Formatter
Reshape one prompt for any model.
Put a line with only --- where the system part ends. Everything above it becomes the system message; the rest is the user message.
About Multi-LLM Prompt Formatter
Multi-LLM Prompt Formatter converts the same prompt between the four layouts different AI products expect. Paste your text once and switch between JSON messages, the array of system and user messages the OpenAI API consumes; XML sections, the tagged blocks Anthropic recommends for Claude; Notes to Markdown, which structures rough notes into a titled, bulleted prompt; and Markdown strip, which removes syntax for models and UIs that prefer plain text. A line containing only --- marks where the system part ends, so role and task stay separated in every output format. Everything converts live in your browser.
How to Use Multi-LLM Prompt Formatter
Paste your prompt
Type or paste the prompt into the left panel. If it has a system part, end that part with a line containing only ---.
Pick the target format
Choose JSON messages, XML sections, Notes to Markdown, or Markdown strip in the toolbar. The hint under the toolbar explains what each mode does with the --- separator.
Copy the result
The converted prompt appears on the right with character and rough token counts. Click Copy or Download to keep it.
Why Use Multi-LLM Prompt Formatter: Common Use Cases
Moving a prompt from a chat UI to API code
A prompt tuned in ChatGPT's web UI becomes a JSON messages array you can paste straight into a fetch call, with the persona as the system message.
Porting prompts between Claude and OpenAI
The same content converts to XML tags for Claude and to JSON messages for OpenAI, so a prompt tested on one provider can be tried on another in seconds.
Cleaning marketing copy out of AI inputs
Markdown strip removes bold, italics, headings, and code fences when a target system renders them as literal symbols.
Structuring meeting notes into a prompt
Notes to Markdown turns the first line into a title, keeps lists, and bolds Label: value lines, turning raw notes into a prompt that reads like a brief.
Multi-LLM Prompt Formatter Specifications
| Input Formats | Text |
|---|---|
| Output Formats | Text (JSON, XML, Markdown, or plain) |
| File Size Limit | No strict limit (dependent on device memory) |
| Processing Engine | 100% Client-side (Runs locally in your browser) |
| Data Retention | Files never leave your device |
| Batch Processing | Single file processing |
Tips for Multi-LLM Prompt Formatter
In JSON mode the output is produced with a real JSON serializer, so quotes and newlines inside your text are escaped correctly every time.
In XML mode, heading lines starting with # become the tag names. Use short headings like # Task and # Context for clean tags.
Build the content with AI Prompt Structurer first, then convert it here when your target product needs another shape.
Validate any JSON you hand-edit afterwards with JSON Formatter before it reaches your code.
Compare two format variants of one prompt with Diff Checker.
Frequently Asked Questions
What is the --- line for?
It marks the boundary between the system part and the user part of your prompt. Everything above it becomes the system message in JSON mode or the system tag in XML mode. Without a --- line the whole input is treated as the user message.
Is the JSON output guaranteed valid?
Yes. The messages array is built with JSON.stringify, not string concatenation, so quotes, backslashes, and line breaks inside your text are escaped by the engine itself.
How do XML tag names get chosen?
Each Markdown heading line becomes a tag: # SQL Teacher turns into sql_teacher. Content without headings is wrapped in a single task tag. Characters that are illegal in tag names are dropped.
Does Markdown strip damage my text?
It removes syntax characters such as #, **, and backticks, and reduces triple blank lines. The words themselves are untouched, but formatting intent is lost, which is the point of the mode.
Can it convert XML or JSON back to plain text?
Yes, indirectly. Paste XML or JSON into the input and use Markdown strip; tags, braces, and quotes remain as plain text you can tidy by hand. The tool does not parse those formats semantically.
Does formatting change how the AI understands my prompt?
It changes the packaging, not the wording. The value is compatibility: each provider recommends its own layout, and this tool keeps one source of truth for the content.