See what your cohort actually asks Claude
Interactive AI logs every prompt, response, and tool call from this workshop straight to a shared dashboard. Participants add one connector to their Claude account. The conversation is unchanged. The visibility is total.
Pick your program
Choose your workshop or cohort from a dropdown. The facilitator added it ahead of time.
Get a personal install link
You enter your email. We mint a token plus an 8-character install code that ties every turn back to you.
Add the connector to Claude
One copy-paste in Claude Code, Claude Desktop, or claude.ai. Authorize. Done — Claude works exactly as before.
Every turn. Every tool call. Every token.
- →The literal prompt you sent — verbatim, full text.
- →Claude’s response — verbatim, full text.
- →Your system prompt and custom instructions.
- →Every MCP / connector / plugin available in the turn.
- →Every tool call: name, arguments, result.
- →Input tokens, output tokens, cache tokens, model, latency.
- →Stable thread id — so multi-turn conversations stay grouped.
Two minutes from first click to logged.
Stop logging at any time by removing the connector — Claude returns to its untraced default. Nothing else on your machine is touched.
On Claude Code, a small plugin hooks every prompt and response and posts them to the backend. On Claude Desktop and claude.ai, an MCP connector handles the OAuth handshake and logs every tool call.
Your participant token is stored as a SHA-256 hash. We can verify it; we can’t recover it. Lose it, re-enroll.