Start here

Organize and understand your data through your AI assistant: give prompts to ChatGPT, Claude, Codex, OpenCode, or Pi, and let the assistant work in your connected workspace. kudhub stores files, data, and websites between conversations; it is not a chat assistant. Its SSH terminal accepts commands, not natural-language prompts.

Connect your assistant

1. Open your invitation link to create an account, then open Agent credentials in account management and create an agent username and password.
2. Choose your client in the table below and connect it through MCP. Check its linked setup guide for requirements. The Agent credentials page shows your deployment's server URL.
3. Give your assistant files or an approved source.

Add a remote Streamable HTTP MCP server at https://manage.kudhub.com/mcp. When the kudhub authorization page opens, enter agent_name@account_handle and its agent password, then approve access. These are separate from the email and password used to manage your account. Use a separate credential per assistant; revoke it when no longer needed.

For OpenCode, the configuration entry is:

{"mcp":{"kudhub":{"type":"remote","url":"https://manage.kudhub.com/mcp"}}}

For Pi with pi-mcp-adapter, use:

{"mcpServers":{"kudhub":{"url":"https://manage.kudhub.com/mcp","auth":"oauth"}}}

After connecting, ask your assistant to list the workspace files. It should discover help, execute, and prepare_upload. Continue giving prompts in that assistant's conversation.

SSH for human access and file transfers

Use SSH from your own terminal to inspect files, run workspace commands, or transfer files. Do not type prompts into SSH. Give those to your AI assistant in its client. Use a credential from Agent credentials and its password:

ssh -p 22 -l 'agent_name@account_handle' kudhub.com

Replace the example username with the one shown on your Agent credentials page.

Try a first task

In your connected assistant's conversation, ask:

“Match these orders, payments, and shipments by order number. Which fully paid orders still need shipping? Save the combined data so we can use it again.”

Include the data, the outcome, and what the assistant may change or share.

Ask a follow-up in the same assistant: “Use the same rules for this new batch.” Files, prepared data, and saved steps remain available for later questions. Updates happen when you ask. Preparation uses data queries and workspace scripts; complex cleanup may need work outside kudhub first.

Use an answer in your assistant, export records as CSV, or ask for a dashboard. For a website, open Website access and select Open staging preview to see the latest completed work on main. The link ends in /_/refs/staging/. To share the result, choose its folder and audience there, then save. Publishing alone does not make a website public.

Find what you need