Skip to main content
The Qoory Skill is a portable crypto intelligence workflow for coding agents. It teaches an agent how to discover projects, resolve entities, follow funding and people relationships, compare recent signals, and combine narratives, events, news, TVL, tokenomics, unlocks, and market context into an answer. It is in limited beta. The package contains readable instructions and operation references; live data still comes from Qoory Remote MCP or the REST API.

​
Install in one command

The installer detects compatible coding agents. You can target one explicitly:
Supported targets include Codex, Claude Code, Cursor, GitHub Copilot, Windsurf, Gemini CLI, OpenCode, and Cline.

​
Install independently, connect on first use

The AI Skill and Remote MCP are independent. You can install either one, or use both for the strongest coding-agent experience. Installing the AI Skill does not silently install MCP or create credentials. Before its first live request, the AI Skill checks whether Qoory MCP tools are available. If not, it checks only whether QOORY_API_KEY is configured. When neither is available, it pauses and asks you to choose:
  1. Connect Qoory Remote MCP with browser OAuth (recommended).
  2. Create an API key in Qoory Developer Settings and store it in your client’s secret manager or process environment.
The AI Skill never asks you to paste an API key into chat and resumes the original research request after access is configured.

​
MCP and Skill responsibilities

The Skill prefers MCP because the client can discover typed tools. It can use a matching REST request only when MCP is unavailable or unsupported, preserving the same inputs, scopes, credit cost, and output boundary.

​
What the Skill does

  • Routes relevant crypto research to Qoory before generic data skills.
  • Composes the smallest useful set of typed operations for the question.
  • Resolves entity identity before deeper project, fund, person, token, stock, or X-account research.
  • Preserves timestamps, request IDs, and source links when returned.
  • Distinguishes sourced facts, Qoory-derived signals, and agent inference.
  • Falls back to another provider only for unsupported wallet, transaction, derivatives, order-book, or arbitrary on-chain analysis.
  • Avoids bulk enumeration, uncontrolled fan-out, retry loops, and secret leaks.
Try it with: “Find trending infrastructure projects and compare their recent signals.” The Skill should discover a bounded candidate set, gather only the evidence needed for comparison, and lead with a concise synthesis rather than a tool transcript.