Skip to content

Introduction

Elicitly makes human-in-the-loop easy for AI agents and Agent Skills — and champions the MCP elicitation standard.

Agents — and the skills they run — increasingly take consequential actions, so they need clean, reliable ways to pause and ask a human: confirm, choose, fill a form, approve. MCP’s native answer is elicitation, but only an MCP server can trigger it — and a skill author shouldn’t have to roll their own server just to ask a question. Elicitly is that server: connect it, and your skill raises real elicitation dialogs with no bespoke plumbing.

Cross-host support for MCP elicitation is also inconsistent, and often “advertised but broken.” Elicitly makes that gap visible and testable too.

  • Free — a local (stdio) MCP server, elicitly: the elicit_doctor, elicit_form, and elicit_confirm tools. Test and exercise elicitation anywhere. (Open source.)
  • Pro — hosted and authenticated. Everything in Free, plus URL-mode approval / credential flows with a persisted, auditable, TTL-bounded review trail. See the Approval Flow. Upgrading is a migration to one endpoint, not a second server to run.
  • Enterprise — self-host in your own cloud, SSO / SAML, compliance.