Skip to content

Features

Spring AI Playground is organized around six product surfaces, designed to be understood in this order. The first four are the author / consume surfaces, Shipyard is the exit that turns what passed into standalone servers, and Observability is the operator surface that watches them all.

  • Tool Studio


    Low-code authoring environment for JavaScript-based tools. Deny-first sandbox, Draft state, MCP server preset catalog, per-tool sandbox capability overrides.

  • MCP Server


    Built-in MCP server over Streamable HTTP, external connections via HTTP / SSE / STDIO / OAuth 2.1, a multi-tab Inspector for tools, resources, prompts, and client primitives, and a preset catalog of 58 preset MCP servers (50 vendor-official remote + 8 community stdio) activatable from the sidebar - see the MCP Catalog directory.

  • Vector Database and RAG


    Both halves of RAG made inspectable: Spring AI's ETL pipeline for ingestion (7 document readers, splitter, metadata enrichers) and Modular RAG for retrieval, authored and tested in Pipeline Studio before chat consumes it.

  • Agentic Chat


    Unified runtime that composes tools and RAG context in one conversational interface - chain workflows and agentic tool-use side by side. Drop a document on the prompt and it becomes conversation knowledge, routed by size and registered in the Vector Database with one click.

  • Shipyard


    Ship what passed as a standalone MCP server: Local Pass tools, proxied tools and RAG knowledge packed into a Docker bundle that runs without the playground, on local or remote Docker, behind the same bearer token. An external AI can do the same over MCP with the saip_shipyard_* management tools; the Shipyard Architecture page explains the bundle and the container.

  • Observability


    Fourteen in-app dashboards across four groups (AI Usage · AI Stack · Runtime · Overview) covering token economics, tool and MCP behavior, RAG quality, host runtime, and a live trace tail - the operator surface that watches the other five, the management tool calls of an external AI included.

The first four surfaces are intentionally connected. A tool authored in Tool Studio is exposed by the built-in MCP server, verified through the MCP Inspector, and consumed by Agentic Chat together with documents indexed in the Vector Database - without restart or redeploy at any step. Every chat turn, tool call, MCP exchange, and vector query is captured by Observability as it happens. The same chain works with an AI at the controls: over the built-in server's management tools, an external client does the authoring, registering, indexing and shipping, and you approve at the gates (AI Operations Architecture).

For a system-level view - runtime layers, data flows, and extension points behind these surfaces - see Architecture.

Further Reading

  • Overview: return to the main product overview and documentation map
  • Getting Started: install the app, configure providers, and choose a runtime
  • Architecture: runtime layers, data flows, and extension points
  • Tutorials: follow end-to-end workflows for tools, MCP, vector search, and agentic chat