Back to directory

SAP A2A Agent Toolkit Plugin

A Claude Code plugin that scaffolds, deploys and connects AI agents to SAP Joule over the A2A protocol on BTP Cloud Foundry, supporting TypeScript agents on Express or CAP and Python agents with LangGraph.

SAP A2A Agent Toolkit Plugin

This plugin adds Claude Code skills and slash commands for building an Agent-to-Agent, A2A, integration between a custom AI agent and SAP's Joule digital assistant. Context-triggered skills scaffold a complete LangGraph agent with A2A support in Express or CAP, generate correct commands for managing BTP subaccounts, entitlements and services through the btp CLI, generate commands for deploying apps and binding services through the cf CLI, and generate commands for compiling and deploying Joule capabilities through the Joule CLI.

User-invoked slash commands scaffold a new agent project with a chosen language, framework and landscape, create the BTP HTTP destination that lets Joule reach the deployed agent, and run a full deployment that pushes the app, wires the destination and deploys the Joule capability in one step. It requires the cf CLI, the Joule CLI, a BTP subaccount with Cloud Foundry runtime, AI Core and Destination services, Node.js for TypeScript agents, Python 3.12 or later for Python agents, and the MTA build tooling for CAP-based agents, and is Apache-2.0 licensed.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.