AI agents

Enterprise agent architecture & platforms

Patterns for agent planning, tool use, human approvals, monitoring and integration with enterprise services, plus the platforms enterprises are deploying agents on.

This page is a catalog of concepts and learning resources. The site does not host a running AI agent or provide a managed service.
๐Ÿค–

Integration troubleshooting agent

Correlate errors across agents, gateways and iFlows, suggest root causes and reference runbooks, with read-only tools and human-approved fixes.

Review safeguards โ†’
โ‡„

Agent-to-agent delegation

Hand work to agents on other platforms with A2A tasks, Agent Cards and push notifications.

Explore โ†’
๐Ÿชช

Agent identity

Delegated versus autonomous agents, token exchange and per-agent credentials.

Explore โ†’
๐Ÿงฉ

Agent orchestration patterns

Human approval gates, task decomposition, workflow state and policy checks.

Explore โ†’
๐Ÿ”Ž

MCP tool discovery

Tools, permission boundaries and filtered discovery for connected agents.

Explore โ†’
๐Ÿ›ก

Agent security risks

The OWASP Agentic Top 10 mapped to controls you can operate.

Explore โ†’

Enterprise agent platforms and protocol support

Where enterprises build and run agents in 2026, and how each connects to the open protocols. Support levels change quickly; verify general availability in each vendor's documentation.

PlatformVendorMCPA2A
Joule Studio / Joule agentsSAPMCP Gateway in Integration Suite for customer tools; SAP-managed MCP for JouleUsed for Joule to discover and delegate to other agents; SAP on A2A steering committee
Copilot Studio ยท Azure AI FoundryMicrosoftSupported; Azure API Management can front MCP serversIntegrated (Linux Foundation, Apr 2026)
Amazon Bedrock AgentCoreAWSAgentCore Gateway exposes tools over MCPSupported through AgentCore Runtime
Vertex AI agent platformGoogleSupportedOriginator of A2A; native support
AgentforceSalesforceSupportedA2A steering committee member
gooseAgentic AI Foundation (from Block)Native MCP clientโ€”