AI Tools
Generate controlled A/B variants of a prompt across the persona, tone/style, output format, and constraints dimensions, then compare them side by side with a heuristic rubric score and a word-level diff against the baseline. A reproducible way to test how a prompt behaves under different framings before picking one for production.
Call this tool from your code in three languages.
curl -X POST 'https://api.elysiatools.com/en/api/tools/ai-prompt-ab-variant-generator' \
-H 'Content-Type: application/json' \
-d '{"basePrompt":"Summarize the attached research paper for a busy product manager. Focus on what changed and why it matters.","dimensions":["persona","tone","format"],"count":4,"renderMode":"cards","temperature":0.5,"useAi":false}'Send a POST request with your inputs as JSON. File parameters require a separate upload first.
POST https://api.elysiatools.com/en/api/tools/ai-prompt-ab-variant-generator| Name | Type | Required | Description |
|---|---|---|---|
| basePrompt | textarea | Yes | The prompt whose variants you want to generate. Its task intent is preserved across variants. |
| dimensions | select | No | Which axes to vary. Defaults to persona, tone, and format. |
| count | number | Yes | How many variants to generate, including baseline A. |
| renderMode | select | No | — |
| temperature | number | No |
Add this tool to your Model Context Protocol server so AI agents can list and call it.
Add this block to your MCP client configuration:
{
"mcpServers": {
"elysiatools-ai-prompt-ab-variant-generator": {
"name": "ai-prompt-ab-variant-generator",
"description": "Generate controlled A/B variants of a prompt across the persona, tone/style, output format, and constraints dimensions, then compare them side by side with a heuristic rubric score and a word-level diff against the baseline. A reproducible way to test how a prompt behaves under different framings before picking one for production.",
"baseUrl": "https://api.elysiatools.com/mcp/sse?toolId=ai-prompt-ab-variant-generator",
"command": "",
"args": [],
"env": {},
"isActive": true,
"type": "sse"
}
}
}After connecting to the SSE endpoint, list the exposed tools:
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/list"
}Invoke the tool by its id, passing arguments built from its parameters:
{
"jsonrpc": "2.0",
"id": 2,
"method": "tools/call",
"params": {
"name": "ai-prompt-ab-variant-generator",
"arguments": {
"basePrompt": "Summarize the attached research paper for a busy product manager. Focus on what changed and why it matters.",
"dimensions": [
"persona",
"tone",
"format"
],
"count": 4,
"renderMode": "cards",
"temperature": 0.5,
"useAi": false
}
}
}| Higher values push variants toward more concrete, specific phrasing. |
| useAi | checkbox | No | When on, rewrite each variant (except baseline A) with DeepSeek (deepseek/deepseek-v4-flash) for production-ready prose. Off = deterministic heuristic output. Needs an API key. |
HTML result
{
"result": "<div>Processed HTML content</div>",
"error": "Error message (optional)",
"message": "Notification message (optional)",
"metadata": {
"key": "value"
}
}Questions or issues? Contact [email protected]