Spring AI Playground
A cross-platform desktop app for building, testing and validating MCP tools in a controlled local sandbox before exposing them to an agent, backed by a built-in MCP server, an agentic chat interface with RAG and human-in-the-loop approval, and detailed observability dashboards.
Spring AI Playground is a desktop application, built with Spring AI, for building, testing and safely running MCP tools in a local environment on macOS, Windows or Linux. Each tool must earn a Local Pass, a local test run against sample arguments, before it is added to the built-in MCP server and becomes callable from the app's Agentic Chat; a tool that has not passed is never exposed to an agent.
The in-process sandbox is described as defense-in-depth for the local build-and-vet loop rather than adversarial-grade isolation, so untrusted tool code is expected to run in container or microVM isolation instead. Built-in observability dashboards cover chat, tokens and cost, models, tool studio, MCP servers and inspector, the vector database, agentic chat, safety, host, Ollama, the web application, logs and traces, backed by a ring buffer with disk persistence.
It ships a bundled catalog of default tools such as web fetch, datetime, math, security, encoding, crypto, filesystem, GitHub, Wikipedia, weather and finance, plus a Korean-domain bundle, and a preset catalog of external MCP servers, and connects to hosts such as Claude Desktop, Claude Code and Cursor.