MCP Integration

Connect your AI agents to any external tool, service, or system — the open extensibility layer that makes OpenRails infinitely expandable

hub

The Extensibility Story

The Model Context Protocol (MCP) is how OpenRails connects AI agents to the outside world. It is an open standard protocol that provides a universal interface between agents and external tools — databases, APIs, file systems, project management platforms, communication tools, and any other service your business relies on.

MCP is not just a technical protocol detail. It is the reason OpenRails can connect to anything. Every built-in integration (Database, Basecamp, Azure DevOps, Teams, OneDrive, and more) uses MCP under the hood. And the same protocol lets you add your own integrations — without waiting for a product update.

Key Value: Your AI agents are never limited to what OpenRails ships out of the box. Any MCP-compliant service immediately becomes a tool your agents can discover and use. And OpenRails' own capabilities become tools that external AI systems can consume. The result: an integration landscape that grows with your needs.

extension

What MCP Means for You

add_circle Add Any Integration

Need to connect to Salesforce, SAP, Jira, Slack, HubSpot, or an internal system? If it has an API, it can be wrapped in an MCP server and made available to your agents. No platform changes required — register the server and your agents discover it automatically.

auto_awesome Zero-Configuration Discovery

Agents do not need to be told about new tools. MCP servers advertise their capabilities — available operations, parameter requirements, and return types. When you add a new MCP server, every agent in the system can immediately discover and use it.

lock Governed and Secure

MCP does not mean "anything goes." Every tool call flows through the OpenRails gateway with authentication, per-tool access control, rate limiting, and full audit logging. Sensitive operations can require human approval before execution.

swap_horiz Bidirectional

OpenRails is both an MCP client (consuming external tools) and an MCP server (exposing its own capabilities). External AI systems can use OpenRails' knowledge base, agents, and data processing as tools in their own workflows.

electrical_services

Built-In Integration Landscape

OpenRails ships with a comprehensive set of production-ready integrations, all powered by MCP:

IntegrationCategoryWhat Agents Can Do
DatabaseDataDirect database queries with approval gates for writes
BasecampProject ManagementAccess projects, message boards, messages, and search (read-only)
Azure DevOpsDevelopmentWork items, sprints, teams, and time logging
EasyProjectsProject ManagementTime tracking, project and task management
Microsoft TeamsCommunicationBot notifications and messages to channels and conversations
OneDrive / SharePointFile StorageFile and folder sync, document retrieval across Microsoft 365
Web SearchResearchInternet search for current information, fact-checking, external context
Web CrawlerContentDeep website crawling with configurable scope, scheduling, and content extraction
Image AnalysisVisionOCR, diagram analysis, chart reading, and visual content interpretation
Site APICustomConfigurable REST API connector for any RESTful service
stars

Platform Capabilities

search Tool Discovery

MCP servers advertise their available tools with descriptions, parameter schemas, and return types. Agents automatically discover what tools are available and how to use them — no hardcoded integrations.

router Gateway & Proxy

OpenRails acts as an MCP gateway, routing tool calls from agents to the appropriate MCP server. The gateway handles authentication, rate limiting, logging, and approval workflows for tool invocations.

approval Approval Workflows

Configure which tool calls require human approval before execution. Critical operations — database writes, external API calls, financial transactions — can be gated with approval workflows that pause agent execution until authorized.

security Security Controls

Per-tool access control lists, parameter validation, and output sanitization. Tools can be restricted by user role, project, or security level. All invocations are logged for audit.

rocket_launch

Expanding the Landscape

OpenRails is designed to connect to any system your organization uses. There are two ways integrations are added:

extension Built-In Connectors

New connectors are added to the platform by the Clarity team. Once a connector is available, it automatically appears in your connector menu — just add your credentials and you're connected. No code, no configuration files, no deployment steps.

Need a connector we don't have yet? Our team can rapidly build new connectors for your specific enterprise systems. The connector framework is standardized, so new integrations are fast to deliver.

hub MCP Tool Servers

For tool-level integrations, register any MCP-compliant server and your agents can immediately discover and use it. The MCP ecosystem is growing rapidly — many enterprise tools already have MCP servers available, and building new ones is straightforward.

External systems can also trigger OpenRails workflows via RESTful API calls, enabling event-driven automation.

Growing ecosystem: The integration library expands continuously. Salesforce, SAP, Jira, Slack, HubSpot, ServiceNow, and other enterprise platforms can all be connected through connectors or MCP — ensuring OpenRails meets your integration requirements as they grow.

architecture

MCP Architecture

AgentMCP Gateway (auth + rate limit + approval) → MCP Server (tool execution) → ResponseAgent

Gateway supports: multiple simultaneous servers | per-tool access control | full audit logging | Human-in-the-Loop approval gates

lightbulb

Use Cases

storage

Database Queries

Agents query production databases using natural language, with approval gates for write operations

bug_report

DevOps Automation

Create work items, update sprint boards, log time, and pull metrics from development tools

api

API Orchestration

Chain multiple API calls across services to fulfill complex requests autonomously

cloud_sync

File Sync & Ingestion

Automatically pull documents from OneDrive, SharePoint, or other file stores into the knowledge base

chat

Team Notifications

Agents deliver alerts, summaries, and reports directly to Microsoft Teams channels

travel_explore

Web Research

Search the internet and crawl websites to gather current information for agent workflows

Related Feature Sheets