MCP Connectors Overview
AxonFlow uses MCP connectors to give AI agents governed access to databases, APIs, SaaS platforms, and storage systems. This is one of the most important product surfaces for teams building serious agent workflows, because it is where data access, permissions, policy enforcement, and auditability meet.
What MCP Means in AxonFlow
In AxonFlow, MCP connectors give you:
- governed access to external systems from AI agents and workflows
- request-side policy enforcement before a connector call runs
- response-side scanning and redaction before data reaches the application or model
- audit trails for connector access across agent workflows, copilots, and RAG pipelines
That makes MCP relevant for more than just "tool use." It is the control point for production AI data access.
Agent Endpoints
The Agent exposes these MCP-related endpoints:
| Endpoint | Purpose |
|---|---|
GET /mcp/connectors | List configured connectors |
GET /mcp/connectors/{name}/health | Check a connector's health |
GET /mcp/health | Aggregate MCP health |
POST /api/v1/mcp/check-input | Policy-only input validation (for external orchestrators) |
POST /api/v1/mcp/check-output | Policy-only output validation (for external orchestrators) |
POST /api/v1/connectors/refresh | Refresh connector config cache |
GET /api/v1/connectors/cache/stats | Inspect cache state |
Connector Types by Edition
Community (base build)
postgres,mysql,mongodb,cassandra,redis,http
Enterprise (enterprise build)
amadeus,s3,azureblob,gcs,salesforce,slack,snowflake,hubspot,jira,servicenow
Connector availability is determined by build tags. The community repo includes only the base connectors. Enterprise builds add the full set.
Why Teams Use MCP Connectors
Common use cases include:
- RAG systems pulling governed context from databases or document stores
- enterprise copilots querying CRM, ticketing, or internal API systems
- multi-agent workflows that need safe access to records, files, and business systems
- regulated AI applications that must redact sensitive data before it reaches the model or user
Request Flow
Every MCP request goes through policy enforcement around the connector call:
Standalone Policy Checks
If your external orchestrator executes connector calls itself, AxonFlow also exposes standalone MCP policy endpoints:
POST /api/v1/mcp/check-inputPOST /api/v1/mcp/check-output
These let you reuse the same MCP policy logic without having AxonFlow execute the connector call directly.
Community, Evaluation, and Enterprise Positioning
- Community is a strong place to start for governed database and API access.
- Evaluation matters when a team is proving real production workflows and needs more realistic limits and deployment confidence.
- Enterprise matters when the connector estate gets broad, multiple teams rely on the platform, and runtime operations, broader connector coverage, and stronger governance guarantees become mandatory.
For the quick-reference view across connector availability, auth posture, and rollout fit, use Connector Capability Matrix.
