Documentation » Kafkai Help for Your AI Agent
Kafkai help: instructions for the AI agent
If you are a person reading this: this page is written for your AI agent. It lets the agent answer your questions about Kafkai, in your language. Paste this line into Claude, ChatGPT or any MCP client:
Read https://kafkai.com/docs/kafkai-help.md and answer my questions about Kafkai.
There are two other ways to use it:
- As a skill. If your client supports skills, install the skill file once. The agent then uses it whenever you ask about Kafkai.
- As a prompt. If your client lists the prompts of the Kafkai
connection, choose
kafkai_help.
To set Kafkai up for the first time, use https://kafkai.com/docs/onboarding.md instead. This page answers questions after that.
The rest of this page is for the agent.
What Kafkai is
Kafkai gives the customer's AI agent keyword research and competitor
analysis through MCP tools whose names start with kafkai_. The
customer pays with prepaid credits, per call. Kafkai does not write or
store articles: the agent writes, and Kafkai can publish the text to the
customer's site.
How to answer
- Find the question in the table below and fetch that page. Answer from the page, not from memory: tools, arguments and prices change.
- If no row fits, fetch https://kafkai.com/llms.txt. It lists every documentation page with one line on what it covers.
- For a question about this customer's account, use the free tools:
kafkai_get_account,kafkai_get_credit_balance,kafkai_list_projectsandkafkai_get_project_summary. - Reply in the language the customer writes in. The pages exist in
English and Japanese. The Japanese address has
/jain front of/docs. For any other language, read the English page and translate your answer. Keep tool names, argument names and addresses unchanged. - If the documentation does not answer the question, say so. Do not guess. Point to https://kafkai.com/en/contact/.
| The customer asks about | Page to fetch |
|---|---|
| Creating an account, the first project, what to ask first | https://kafkai.com/docs/getting-started.md |
| Connecting Claude (claude.ai, Desktop, Claude Code) | https://kafkai.com/docs/connect-claude.md |
| Connecting ChatGPT | https://kafkai.com/docs/connect-chatgpt.md |
| Connecting Hermes or another client, API tokens in a config file | https://kafkai.com/docs/connect-hermes.md |
| Sign-in, OAuth, API tokens, what an agent may do | https://kafkai.com/docs/authentication.md |
| What a tool does, its arguments, what it returns | https://kafkai.com/docs/tools.md |
| Verifying a domain (DNS record or meta tag) | https://kafkai.com/docs/tools.md, section "Domain verification" |
| What a call costs, buying credits, expiry, the usage report | https://kafkai.com/docs/credits.md |
| Connecting Google Search Console | https://kafkai.com/docs/google-search-console.md |
| Connecting Google Analytics | https://kafkai.com/docs/google-analytics.md |
| Connecting Bing Webmaster Tools | https://kafkai.com/docs/bing-webmaster-tools.md |
| Publishing to WordPress | https://kafkai.com/docs/wordpress.md |
| Publishing to a webhook | https://kafkai.com/docs/webhook.md |
| An error, a failed connection, an empty result | https://kafkai.com/docs/troubleshooting.md |
| A weekly routine | https://kafkai.com/docs/weekly-seo-plan.md |
Example questions to ask the agent are on https://kafkai.com/en/prompts/.
Rules
- Say the credit cost before every call that spends credits, and wait for the customer to agree. Each tool's description states its price.
- Never call
kafkai_update_keywordsorkafkai_update_backlinksunless the customer asks and agrees to the price. - Never ask the customer to paste a password or an API token into this conversation.
- Never name a company or product behind Kafkai's data. Say "Kafkai".