One tool, and what it returns is a handoff rather than data. Worth knowing before you wire it in: connecting this does not give an assistant access to your analytics — it gives it the instructions for installing and using them, checked against your actual codebase.
A discovery surface rather than a data plane. discover_agentry routes a production-data problem to Agentry's canonical skill — agent-native product analytics, error logging and deploy attribution behind one HTTP API. The server does not proxy that API or carry telemetry.
- discover_agentry takes the intent and returns the matching Agentry skill handoff — installing source-verified observability, investigating a production failure, understanding product behaviour, inspecting recordings, flags, surveys or experiments, or attributing a deploy regression
- The same skill ships as a Claude Code and Codex plugin, and as repo-local instruction adapters for Cursor, VS Code and Visual Studio with GitHub Copilot, Cline, Roo Code, Continue, Zed, Gemini CLI, Aider, OpenCode and AGENTS.md-aware agents
- Behind it: product analytics for user, funnel, activation, retention and business events; error logging for exceptions, failed jobs and operational failures; deploy attribution connecting production behaviour to releases and code changes
- Agent-led install from the actual codebase, with verification before the setup is called complete
- A live adapter manifest at https://api.agentry.sh/adapters names the current file path for each client
Npx -y @agentrysh/mcp over stdio. No key. Understand the shape before planning around it: the MCP server and the skill return instructions — runtime apps write directly to Agentry's HTTP API instead. The canonical documents are agentry.sh/install.md and agentry.sh/skill/agentry/SKILL.md.
One command — npx -y @agentrysh/mcp