Call A Human FAQs

Question 1

Is there an audit trail for human interactions with AI agents?

Accepted Answer

Absolutely. Call A Human maintains a comprehensive, append-only JSONL audit log that records all human interaction events. This provides a clear, unalterable record for transparency, accountability, and debugging purposes.

Question 2

What is Call A Human?

Accepted Answer

Call A Human is an MCP server that equips AI agents with a 'pause button', enabling them to solicit human approval or crucial information before executing critical, sensitive, or irreversible actions. It ensures human oversight in AI workflows.

Question 3

How does Call A Human prevent irreversible AI actions?

Accepted Answer

AI agents can utilize the `request_approval` tool to prompt a human for a decision before proceeding with sensitive operations. The AI agent's process is blocked until human approval or denial is received, safeguarding against unintended consequences.

Question 4

Which AI clients and notification channels are supported?

Accepted Answer

Call A Human is compatible with MCP-enabled AI clients such as Claude Desktop, Cursor, and Windsurf. It supports multiple notification channels including Slack, Telegram, and native macOS system dialogs for flexible deployment.

Question 5

Can AI agents gather information from humans using this tool?

Accepted Answer

Yes, AI agents can employ the `ask_human` tool to pose questions directly to humans. This allows them to receive text replies, gather necessary context, or obtain specific data points to inform their subsequent actions and decision-making.

Call A Human

Call A Human

主な機能

ユースケース

主な機能

ユースケース