Skip to Content

MCP Asset Setup, Knowledge RAG & Agent/Multi-Agent Orchestration

In the isA Platform, an enterprise AI agent is not an isolated prompt, but a sophisticated system assembled from model gateways, MCP toolkits, skills, Knowledge Vector Stores, and multi-agent topologies.

This chapter demonstrates how to generate and configure MCP assets, mount enterprise vector stores, assemble a Single ReAct Agent, and design a Fixed Multi-Agent Team Canvas on https://isagent.io/console.

Complete Walkthrough Video

Full-Stack Agent & MCP Studio (1m 07s)

Console Deep Dive: MCP Assets, Vector Knowledge & Multi-Agent Studio

Open Agent Studio
Jump to Key Action
Comprehensive tour of enterprise agent orchestration: inspecting 296 MCP tools, mounting hybrid vector stores, assembling Single ReAct agents with fine-grained tool bindings, orchestrating multi-agent teams on the visual Team Canvas, and auditing OpenTelemetry trace waterfalls.

Core Operational Stages

  1. MCP Capability Hub: Inspect and publish tools, skills, prompts, and remote MCP servers.
  2. Knowledge & Vector Stores: Ingest documents (Feishu, S3, PDF), calculate embeddings, and configure hybrid RAG retrieval.
  3. Single ReAct Agent: Select models, customize system instructions, and bind specific MCP tool clusters and skills.
  4. Multi-Agent Team Canvas: Build specialized teams (Research, Code, QA, Chat) with entry routing and automated dependency validation.
  5. Observability & Traces: Run test sessions in Playground and inspect real-time OpenTelemetry trace waterfalls.