Reference file

claygent-guide.md

claygent-guide-md.md

Claygent Guide

What is Claygent?

AI research agent that combines Google Search, ChatGPT, and web scraping into a single tool. It browses the web, finds public data, and reports results. Use Claygent when data doesn't exist in standard databases and you need real-time, nuanced information from the web.

How to Access

  1. Open a table
  2. Go to the Enrichment panel (right side)
  3. Select "Claygent" under the AI section

Model Selection

Model When to Use Cost
Claygent Neon 90% of tasks — extraction, formatting, simple lookups Lowest
GPT-4 Mini Slightly more complex reasoning, still cheap Low
GPT-4 Deep reasoning, multi-step analysis Higher
Claude Opus Deep reasoning, nuanced interpretation Higher

Rule: Use Neon or GPT-4 Mini for everything. Advanced models are benchmarked for solving calculus — you just want to say "I saw on your website you help fitness enthusiasts breathe better." Fix problems with better prompts, not more expensive models.

Credit Cost

  • Simple questions (yes/no, single lookup): 1 credit
  • Deeper analyses (multi-page scraping, reasoning): 2-3 credits
  • Always use conditional runs to avoid wasting credits on empty rows

Claygent vs Standard Enrichments

Use Case Use This
Email/phone/company data Standard enrichments (Apollo, Prospeo, etc.)
Data exists in a database Standard enrichments
Data doesn't exist in databases Claygent
Need real-time web context Claygent
Very specific/nuanced information Claygent
Scraping a specific webpage Claygent

8 Prompting Rules (Eric Noski)

Rule 1: The 10-Minute Manual Research Rule

Ask yourself: "What would I do if I researched this person for 10 minutes?" What information would I look for? THEN automate that with Claygent. Never guess what you could do — start from manual research.

Rule 2: One Task Per Claygent

Bad: Claygent that checks site content + decides if e-commerce + classifies industry Good: Claygent #1: Retrieve site content → Use AI #2: Is it a CPG? → Use AI #3: What products? → Use AI #4: What industry?

AI fails when given too many tasks. GPT-4 Mini costs almost nothing — separate the tasks.

Rule 3: Use "purple" as Null Keyword

When nothing is found, output a specific keyword like "purple" instead of letting AI say "No information found" in 100 different ways. Makes downstream filtering easy.

If no case study found, output 'purple'

Rule 4: Always Include Examples in Prompts

Even a mediocre prompt becomes excellent with 3-4 examples. Manually research a few cases, include them in the prompt. Shows AI exactly what you want.

Rule 5: Use the Metaprompter

  1. Go to ChatGPT
  2. Explain what you want to accomplish
  3. Ask: "Create a prompt for this"
  4. Ask: "What questions do you have to improve this prompt?"
  5. Answer the questions → improved prompt

Rule 6: Ask for Reasoning Before the Answer

"Tell me the reasoning why you think it's this kind of company before you give me the final answer." AI predicts the next word — by explaining first, it has more context to answer better.

Rule 7: Prepare for Edge Cases

20% of the work = putting integrations together. 80% = adjusting the AI. Review outputs constantly — as soon as an output doesn't look right, adjust the prompt.

Rule 8: Iterate Without Spending Credits

Use the builder to test and iterate on prompts before running on the full table. Test on 1-3 rows in the builder until the output is exactly what you want.

Prompt Template

Visit {{company_url}}.

[Specific instruction — one task only]

Return format: [text/number/URL/true-false]

If not found, output 'purple'.

Examples:
- Input: Acme Corp → Output: "Series B, $25M, 2024"
- Input: Widget Inc → Output: "purple"

Output Formatting

Available formats: text, number, URL, true/false, custom

Always specify the exact format you want in the prompt to avoid inconsistent outputs.

Frontal Production Prompts

Company Qualification

You are qualifying companies for a B2B outbound campaign. Based on the company information provided, assign a tier:

Tier 1: B2B software companies with 40+ headcount that sell to other businesses
Tier 2: B2B companies that sell digitally but are not software companies
Tier 3: Companies in a relevant industry but not our ideal fit

Company: {{Company Name}}
Industry: {{Industry}}
Employee Count: {{Employee Count}}
Description: {{Company Description}}

Return ONLY: "Tier 1", "Tier 2", or "Tier 3" followed by a one-sentence reason.

Contact Qualification

Based on the following job title, assign a seniority tier:

Tier 1: C-level, VP, Founder, Head of, Director (decision-makers with budget authority)
Tier 2: Manager, Senior (influencers who can champion internally)
Tier 3: Associate, Coordinator, Specialist (end users, not decision-makers)

Title: {{Title}}

Return ONLY: "Tier 1", "Tier 2", or "Tier 3"

Personalized Opening Line

Write a cold email opening line (1 sentence, under 20 words) based on this signal:

Signal: {{Signal/Trigger}}
Company: {{Company Name}}
Person: {{First Name}}, {{Title}}

Rules:
- Reference the signal naturally, don't force it
- Sound like a human, not AI
- Don't pitch anything yet
- Don't start with "I noticed" or "I saw"
- Make it about THEM, not you

Tech Stack Detection (0 credits via GPT-4o-mini API)

Visit this BuiltWith URL and list the key technologies used by this company:
{{BuiltWith URL}}

Return a comma-separated list of the top 10 most relevant technologies (CRM, marketing automation, analytics, web framework only). Ignore tracking pixels and common libraries.

Examples

Example 1: Detect free trial availability

Visit {{company_url}}. Does this company offer a free trial?
Look for: "free trial", "try for free", "start free", "demo", "trial period".
Return 'Yes' if found, 'No' if not found, 'purple' if the page cannot be loaded.

Example 2: Extract case study industries

Visit {{company_url}}/customers or {{company_url}}/case-studies.
List the industries of the first 3 case studies shown.
Format: "Industry1, Industry2, Industry3"
If no case studies page exists, output 'purple'.

Example 3: Check for competitor mentions

Search Google for: site:{{company_domain}} "{{competitor_name}}"
Does this company mention {{competitor_name}} on their website?
Return 'Yes' if found with the specific page URL, 'No' if not found.

Example 4: Find hiring signals from careers page

Visit {{company_url}}/careers or {{company_url}}/jobs.
Count the number of open roles. List the departments hiring.
Format: "X open roles — Engineering, Sales, Marketing"
If no careers page exists, output 'purple'.