MCP Agent Hub
Give AI modular tools and data connectors without rebuilding the model layer.
Interactive platform map
Architecture in context
The focused blueprint, its required foundation, and declared recommendations.
MCP Agent Hub
Give AI modular tools and data connectors without rebuilding the model layer.
What this replaces
Chat interfaces answer questions but cannot safely interact with databases, crawlers, reports, documents and operational systems.
What your team gains
A standardized capability layer where MCP-capable clients can discover tools, resources and prompts and invoke permitted actions.
What is inside the blueprint
How it fits the platform
AI client -> MCP protocol -> specialized MCP servers -> external systems/data/tools -> structured results back to the model.
neo4j-mcp-server, session-recall-mcp, scrapy-mcp, google-docs-mcp, grafana-reports, MCP-compatible clients
- AICORTEX Core PlatformMCP servers are Core-managed containers behind Core's ingress.
From prerequisites to operation
- MCP-capable client
- Tool-specific credentials
- Authorization and risk policy
- Deploy/register MCP servers
- Choose local stdio vs remote Streamable HTTP
- Apply auth and network controls
- Define tool scopes
- Test client discovery/invocation
- Tool descriptions/schemas
- Permissions
- Transport/auth
- Resource namespaces
- Rate limits
- Approval requirements
- Connector/API drift
- SDK compatibility
- Credential rotation
- Tool audit logs
- Dependency pinning
What this unlocks with other layers
Controlled Agency
Agents can act across systems while risky writes and commands remain human-gated.
AI SRE Foundation
Agents can reason over the same telemetry and history humans use for diagnosis.