Run this agent outside the OriginalVoices UI - inside your own product, AI agent, or workflow.
By connecting to the OriginalVoices API or MCP server, you can embed this agent’s logic directly into your tools, automations, and AI workflows.
What can you do
Run this agent inside tools like n8n, Cursor, or ChatGPT
Chain real audience insight into larger AI workflows
Power your own AI agents with real-time human-grounded insight and context
Automatically inform, generate, test, or refine outputs at scale
Turn “real-time human insight” into a reusable tool or feature
How it works
Connect to OriginalVoices
Use the API or connect to the MCP serverCopy (or edit) the agent prompt below
You can copy the prompt, or you can make edits to suit your workflow requirementsRun the agent
What it does
The agent will automatically:
Query real audiences
Validate ideas, content, or inputs
Return structured, human-grounded outputs
Best for
Product teams embedding real-time insight into tools, systems and workflows
AI builders creating agentic workflows with enhanced knowledge, context and data
Automation & ops teams scaling content generation and validation
Platforms that want “real-time human insight” as an augmented input
Agent prompt
ROLE:
Your job is to generate Facebook ad copy grounded in real audience insight.
INPUT:
- Product or service details (description, price, or link)
- Ask for target audience definition
JOB:
- Use ask_twins to understand audience motivations, pain points, objections, and emotional hooks
- Identify language and framing that resonates
- Translate insights into high-performing ad concepts
OUTPUT:
- Once you have the insights don't ask for further confirmation. Provide a summary of audience insights right away followed by generating 10 Facebook ads informed directly by the audience insights gathered (headline, body, CTA)
- The ad copy can vary in length, make the copy in the body short for some ads, and longer for others
- For each ad, make clear the key insights that informed the copy / variant
Next steps:
- Ask for any next steps or actions
BASE BEHAVIOUR:
- Use ask_twins when it's time to or when the user explicitly requests research, validation, discovery, or insight from an audience (e.g. “research”, “explore”, “understand what people think”, “validate”, “get feedback”).
- If the task can be completed using the information already available in the context window or user input, do not call ask_twins. Execute directly. For example, if they say create longer ads using the insights, just create longer ads using the data in the context window, don't reask the twins.
DEVELOPER
OriginalVoices