@delvry/mcp

MCP server for Delvry — agent-friendly Q&A surveys.

Install

Claude Code

claude mcp add delvry -- npx -y @delvry/mcp

Claude Desktop

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "delvry": { "command": "npx", "args": ["-y", "@delvry/mcp"] }
  }
}

Claude.ai (web Connectors)

Settings → Connectors → Add custom → URL: https://delvry.ai/mcp

(For self-hosted Delvry, use your own URL.)

Bootstrap an account

The first time the agent calls delvry_signup, the server creates an account and persists the API key to ~/.delvry/config.json. Subsequent calls use it automatically — no restart required.

Tools (11)

delvry_signup, delvry_create_survey, delvry_ask, delvry_get_responses, delvry_await_responses, delvry_get_recipients, delvry_close_survey, delvry_list_surveys, delvry_validate_survey, delvry_summarize_responses, delvry_delete_response.

What's new in 1.7.0

delvry_converge gained a seed-engine workflow (new params, not new tools):

• seed_beliefs — assert your current definition per term so recipients confirm/correct instead of authoring from blank (an agent belief is a hypothesis; a human still confirms). Already-settled terms are skipped.
• One consolidated message per recipient (was one email per term).
• nudge_after — automatic follow-ups (default 3d/7d).
• Per-term routing — an audience_hint on a seed belief plus a recipient role_hint route a term only to the people who own it.

Configuration

| Env var | Purpose | Default | |---|---|---| | DELVRY_API_KEY | Your account's API key | falls back to ~/.delvry/config.json | | DELVRY_ENDPOINT | Delvry instance URL | https://delvry.ai |

Docs

• https://delvry.ai — marketing + survey schema
• https://delvry.ai/llms.txt — agent-readable install + usage doc
• https://delvry.ai/skill.md — bundled skill markdown