SuperAPI AI Directory

elicitly

🎖️ 📇 🏠 🍎 🪟 🐧 - Human-in-the-loop via MCP elicitation: native confirmation dialogs and typed forms from JSON schemas in the host, plus a check of supported features.

Open website

About this tool

Elicitly Free Edition is used to provide human-in-the-loop capabilities for AI prompts and Agent Skills in the MCP (Model Context Protocol) ecosystem. The problem it solves is: although MCP natively supports asking humans questions via elicitation, the level of support across MCP hosts is inconsistent, and only MCP servers can trigger this capability. Elicitly is exactly such a server, or it can be embedded into your own server as a tool library via @elicitly/tools.

Main capabilities

  • Local MCP server: can be run directly from npm via npx -y elicitly (stdio mode), with no need to clone code or perform additional configuration.
  • Embeddable tool library: @elicitly/tools allows human-in-the-loop capabilities to be integrated into custom MCP servers.
  • Interaction tools: provide the Agent with the following calls:
    • elicit_confirm: a native confirmation dialog (for example "Ship it?").
    • elicit_form: generates a typed form based on a user-provided JSON Schema.
    • elicit_doctor: detects whether the current MCP host truly supports elicitation.
    • elicit_display: renders LiquidJS templates as an inline MCP App for display.
  • Host support diagnostics: the output of elicit_doctor feeds into the public Support Matrix.

Usage

  1. Run the local stdio MCP server via the command line: npx -y elicitly.
  2. Register the server in any MCP host (the documentation uses Claude Code as an example; other stdio hosts are equally applicable).
  3. In Claude Desktop, the command line can be bypassed: download elicitly.mcpb from the latest release and open it to complete a one-click, self-contained desktop extension installation.
  4. After installation, have the Agent call elicit_confirm, elicit_form, elicit_doctor, or elicit_display.

Applicable scenarios

  • Need to request user confirmation before AI performs critical operations (such as publishing, deployment, etc.).
  • Need AI to collect structured information from users (generate forms via custom JSON Schema).
  • Need to verify or understand the actual support for elicitation by a given MCP host.
  • Need to integrate human-in-the-loop capabilities into a self-built MCP server.

Limitations and version notes

  • This repository is the open-source free edition, including the local MCP server and the @elicitly/tools library.
  • The hosted Pro edition (providing persistent approval and review pages, audit trails, and organizational dashboards) is located at elicitly.ai and is not part of this repository.
  • More details regarding pricing, charges, platform compatibility, etc. are not currently provided in the available materials.