Quick Links
Connect with us
AI Agent Hub
Govern, orchestrate, and automate all your AI Agents
Overview of AI Agent Hub
AI Agent Hub is a centralised enterprise platform that manages, orchestrates, and governs all AI agents across your organisation. It serves as a secure proxy: every call between users and agents flows through the Hub, ensuring full visibility, security, and control at every step.
Whether you are onboarding a Quick Agent using pre-built MCP tools in minutes, or registering an externally-built Custom Agent from any LLM framework, AI Agent Hub delivers the same governed, audited, and zero-trust execution environment across both agent types.
Core features of AI Agent Hub
Centralised agent registry
One catalogue for all Quick and Custom agents across the organisation. Filter by technology, category, type, and mode with full-text search.
Zero-trust credential management
Credentials stored in Vault, injected at execution time only. The LLM never sees credentials, tool names, or raw parameters.
Dynamic human-in-the-loop
Per-operation HITL with group-based approvals. Execution pauses automatically, letting you approve or reject critical actions directly in the chat thread.
106+ pre-built MCP tools
A curated tool catalogue for enterprise infrastructure: Kubernetes (52), SaltStack (42), and RPA (12), with more platforms on the roadmap.
Multi-dimensional RBAC
Subscribe agents by org hierarchy, project, user role, or user group, or combine all four dimensions for more granular access control.
Immutable audit trail
100% of executions captured as 8 distinct audit event types with step-level timeline, per-operation detail, and CSV export for compliance.
What you get with AI Agent Hub
FAQs about AI Agent Hub
What is the difference between a Quick Agent and a Custom Agent?
Quick Agents are built and managed entirely within AI Agent Hub using pre-built Model Context Protocol (MCP) server tools. They require no endpoint configuration and go live immediately upon creation — no approval step is needed. Custom Agents are built externally using any large language model (LLM) framework such as LangChain, AutoGen, or custom code, and registered in the Hub by providing their invocation endpoint URL. Custom Agents require Admin approval before becoming active. Both types receive the same role-based access control (RBAC), human-in-the-loop (HITL), credential management, and audit trail treatment from the Hub.
How does Human-in-the-Loop (HITL) work in AI Agent Hub?
HITL can be configured at the agent level or per individual operation. When the agent reaches a HITL-gated operation, execution pauses automatically and an approval card appears inline in the chat thread. The configured approver group is notified and can approve or reject the action directly in the interface. Key points:
- Selecting Approve allows the agent to execute the gated operation and continue
- Selecting Reject cancels the operation — no changes are made to any backend system
- Both outcomes are logged as immutable audit events (HITL_APPROVED / HITL_REJECTED)
- HITL is recommended for all write, delete, and irreversible operations
What MCP tools are available out of the box?
AI Agent Hub ships with 106+ pre-built tools across three MCP server platforms:
- Kubernetes MCP Server — 52 tools covering cluster health, workloads, pods, Helm, ingress, services, CRDs, and YAML apply/patch
- SaltStack MCP Server — 42 tools for minion management, state execution, job management, Kafka, RabbitMQ, and agentless SSH/WinRM
- RPA MCP Server — 12 tools for executing automation sequences, journal management, milestone tracking, and business process automation
Additional MCP servers for databases, networking, cloud providers, and ITSM are on the Phase 2 and Phase 3 roadmap.
How is access to agents controlled?
Agent access is managed through subscriptions configured by Agent Admins or Creators. A subscription grants access based on one or more of four criteria, any of which can be combined:
- Application Hierarchy — org path (Unit → BU → Domain → Platform → Application → Environment)
- Service Hierarchy — technical org path (Unit → Ministry → Department → Organisation → Application)
- Project — all members of a named project regardless of hierarchy position
- User Role or User Group — users who hold a specific job function role or are members of a named group
Users matching any criterion gain access immediately. Subscription changes, including revocations, take effect instantly.
Is AI Agent Hub compliant with enterprise audit requirements?
Yes. Executions are stored as immutable audit events: 8 types covering runs, agent lifecycle, subscriptions, human-in-the-loop decisions, and configuration changes. Each record captures actor, asset, operation, outcome, and timestamps at session and step level. No separate logging stack is required.
Related Products
Application Performance Management
Application Performance Management
End-to-end visibility into application performance.
Log Analysis
Log Analysis
Centralised, scalable log management and analysis for actionable cloud insights.
SIEM
SIEM
End-to-end visibility into security events across IT and cloud ecosystems.
Ready to build smarter infrastructure?
Share a few details and our team will reach out with product details, pricing, and a demo.
By selecting 'Submit', you authorise Jio Platforms Limited to store your contact details for further communication.