Skip to Content

Console Operations Guide

The AIX Enterprise Console provides a unified operational command center for multi-tenant agent hosting, model routing gateways, MCP tool cluster integration, and end-to-end observability governance. This guide walks enterprise developers and IT architects through workspace configuration, Agent Studio asset management, Playground thought trace debugging, remote MCP tool cluster integration, and production API key RBAC scoping.

Console Walkthrough (26s)

Live Walkthrough: Orchestrate, Test & Deploy an AI Agent

Open Live Console
Jump to Key Action
Real-world engineering walkthrough: navigating 474 production agents, inspecting multi-tenant governance, editing agent configurations in Agent Studio with 377 MCP tools, and collaborating with live Copilot (Qwen 3.8).

5 Core Operational Phases

Phase 1: Workspace Overview & Operational Pulse

After logging into the enterprise production console (https://aix.snnc.cc/console), navigate to Workspace Overview:

  • Operational Pulse: Observe real-time API call volumes, Token throughput peaks (TPS), end-to-end P99 latency, and active long-running agent workflows across your organization.
  • Workspace Readiness: Verify health statuses for multi-tenant isolation boundaries, private Qwen model gateway nodes, and 377+ MCP tool cluster endpoints.
  • Quota & Budget Protection: Monitor monthly Token consumption budgets, rate limiting triggers, and automated circuit breaker thresholds.

Phase 2: Agent Studio Orchestration & Lifecycle Management

Navigate to Workspace → Agents (/dashboard/agents) to inspect and manage production agents. Click + Create Agent:

  1. Metadata & Identifier: Define a globally unique agent slug and business alias (e.g., sweetnight-pricing-review-agent).
  2. Base Model Selection & Inference Parameters: Select appropriate foundation models (such as privately hosted Qwen3-72B, DeepSeek-V3), configuring Temperature and Top-P bounds.
  3. System Prompt & Guardrails: 
    You are a pricing compliance and elasticity review agent for SweetNight cross-border commerce.
Adhere strictly to gross margin baselines and ODCS data contract specifications.
Trigger MAD anomaly attribution whenever sales volatility exceeds threshold limits.
  4. Skills & Memory Wiring: Configure short-term conversation sliding windows, vector store retrieval scopes, and specialized enterprise skill packages.
  5. Release & Snapshot Tags: Save drafts, run canary testing, and promote to an immutable release tag.

Phase 3: Playground Interactive Testing & Thought Trace Debugging

Navigate to Workspace → Playground (/dashboard/playground) to test agent behavior within a sandboxed environment:

  • Dynamic Agent Switching: Switch between agent versions and underlying models to perform live A/B comparisons.
  • Thought & Context Trace: Expand the inspection drawer to view dynamic prompt assembly, Chain-of-Thought reasoning steps, and tool execution inputs/outputs.
  • Latency Decomposition: Isolate prompt compilation overhead, time-to-first-token (TTFT), and remote MCP network roundtrips.

Phase 4: Mounting Enterprise MCP Tool Clusters

Navigate to Workspace → MCP → Servers (/dashboard/mcp/servers):

  • Model Context Protocol (MCP): Standardizes agent connectivity across heterogeneous enterprise backends (ERP, MES, StarRocks, Feishu).
  • Connect MCP Servers: Register enterprise endpoints via Server-Sent Events (SSE) or secure HTTP proxies.
  • SHA-256 Contract Verification: Inspect tool schema specifications, parameter contracts, and cryptographic digests to prevent tampering.
  • Tool Assignment: Link verified MCP tools to agents in Agent Studio, granting autonomous access to backend business operations.

Phase 5: Production API Key Governance & Scoped RBAC

Navigate to Developer Settings → API Keys (/dashboard/settings/api-keys):

  • Issuance: Click Generate New Key to provision dedicated credentials for microservices or pipelines.
  • Least-Privilege Scopes: Restrict key capabilities (e.g., read-only execution agent:execute, forbidding configuration updates agent:write or privileged tool escalation mcp:admin).
  • Cryptographic Storage: Private tokens are displayed once upon generation and persisted using Argon2/SHA-256 salted hashes.
  • Production cURL Example: 
    curl -X POST https://aix.snnc.cc/api/v1/chat/completions \
  -H "Authorization: Bearer aix_live_your_token_here" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-72b",
    "messages": [{"role": "user", "content": "Analyze top return drivers for the mattress line"}]
  }'

Next Steps