The Legacy Protocols Your AI Agent Can't Speak. We Translate.
PipeGears is the governed MCP gateway that exposes SQL databases, SFTP file systems, SOAP/XML services, EDI X12, MQTT brokers, and on-prem systems as safe, audited tools for Claude, ChatGPT, Copilot, and any MCP-compatible agent.
{
"tools": [
{
"name": "sql.query",
"description": "Query SQL databases (SQL Server, MySQL, PostgreSQL)"
},
{
"name": "sftp.read",
"description": "Read files from SFTP servers"
},
{
"name": "edi.parse",
"description": "Parse EDI X12 documents (850, 810, 856)"
},
{
"name": "soap.call",
"description": "Call SOAP/XML web services"
},
{
"name": "mqtt.publish",
"description": "Publish to MQTT topics"
},
{
"name": "flow.execute",
"description": "Run a pre-built integration flow"
}
]
}Your AI Agent Has a Legacy Problem
Your AI agent can search the web, summarize documents, and write code. But it can't query a SQL Server behind a firewall. It can't read EDI files from an SFTP server. It can't call a SOAP service or publish to an MQTT broker.
The MCP ecosystem has exploded: 10,000+ servers published, hundreds of millions in venture funding, thousands of connectors for modern cloud SaaS. None of them reach the systems that actually run manufacturing.
We built PipeGears for those systems.
MCP Tools for the Systems That Matter
sql.query / sql.execute
Query and write to SQL databases: SQL Server, MySQL, PostgreSQL, and more. Connected through the Bridge Agent, your on-prem database becomes AI-accessible without exposing it to the internet.
sftp.list / sftp.read / sftp.write
List, read, and write files on SFTP servers. Your EDI mailbox, supplier data feeds, and file-based integrations, all available to your AI agent.
rest.call
Call any REST API with full OAuth 2.1 authentication. Shopify, Zoho, QuickBooks, HubSpot; any system with a REST API.
soap.call
Call legacy SOAP/XML web services. Your agent sends JSON, PipeGears handles the WSDL, XML envelope, and namespace translation. Your AI agent never touches XML.
mqtt.publish / mqtt.subscribe
Publish and subscribe to MQTT topics. Connect AI agents to factory-floor devices, sensors, and industrial IoT systems.
edi.parse / edi.build
Parse and generate EDI X12 documents: 850 (Purchase Order), 810 (Invoice), 856 (Ship Notice). The only MCP server in the ecosystem that speaks EDI.
flow.execute
"Sync my Shopify inventory with the ERP" becomes one command. Run a complete, pre-built PipeGears integration flow as a single tool call.
engine.status
Check integration health metrics: flow execution counts, error rates, last successful runs. Your AI agent can monitor your integrations and alert you to problems.
What You Get vs. a Raw MCP Server
| Capability | Raw MCP Server | PipeGears MCP |
|---|---|---|
| Authentication | Maybe an API key | OAuth 2.1 + PKCE (spec-compliant) |
| Multi-tenant isolation | Single-tenant | Per-workspace isolation |
| Retry / reliability | None | Circuit breaker, exponential backoff |
| Deduplication | None | Built-in (unique in the MCP ecosystem) |
| Rate limiting | None | I/O Governance |
| Credential storage | Cleartext config | Encrypted per-user key chain |
| On-prem access | None | Bridge Agent reaches behind firewalls |
| Large file handling | Crashes | Content system with disk spillover |
| Data transformation | None | Full transformation pipeline |
| Audit trail | None | Workspace metrics + governance API |
Anyone can build an MCP server that queries a database. PipeGears queries a database behind a firewall, through a bridge agent, with retry logic, deduplication, rate limiting, and encrypted credentials, in a multi-tenant, production-grade environment.
Connect in 4 Steps
Sign Up
Create your PipeGears account and get your MCP endpoint URL.
Configure Your MCP Client
Add PipeGears to your AI tool's MCP configuration.
{
"mcpServers": {
"pipegears": {
"url": "https://mcp.pipegears.com/ws/your-workspace-id",
"auth": "oauth2"
}
}
}Connect Your Data Sources
Configure your SQL databases, SFTP servers, APIs, and other systems in the PipeGears dashboard. On-prem? Deploy a Bridge Agent.
Ask Your AI
Your AI agent discovers PipeGears tools automatically via MCP. Start asking: "What orders are pending?" "Sync my inventory." "Parse the latest EDI file."
What It Looks Like in Practice
Ask Claude About Your Factory
"How many orders are pending for this week?"
You have 23 pending orders this week, totaling $47,200. The largest is order #4521 for Riverside Kitchen & Bath at $8,400.
↳ pipegears.sql.query → WEB-CAB production database
"What's the status of order #4521?"
Order #4521 is in CNC cutting. Estimated completion: Thursday.
↳ pipegears.sql.query → WEB-CAB production database
Sync My Inventory Now
"Sync my Shopify inventory with the ERP right now."
Done. 342 products synced. 3 products had stock discrepancies: Shopify showed 0 for SKUs PKG-2240, PKG-2241, and PKG-2255, but the ERP shows 12, 8, and 15 respectively. Updated on Shopify.
↳ pipegears.flow.execute → inventory-sync flow
"Why were those three wrong?"
These SKUs had a manual inventory adjustment in the ERP yesterday at 3:47 PM, after the last scheduled sync.
↳ pipegears.sql.query → ERP audit table
Process the EDI File
"Check if we got a new PO from Home Depot."
Yes. EDI 850 received today at 6:12 AM via SFTP. Purchase order for 240 units across 8 SKUs, total value $18,600. Ship-by date: April 18.
↳ pipegears.sftp.read → EDI mailbox pipegears.edi.parse → EDI X12 850
"Create the production order from this PO."
Production order #7844 created in the ERP. 8 line items, priority: standard.
↳ pipegears.sql.execute → ERP production orders table
Built for the Systems Others Can't Reach
You are here. Nobody else is.
Cloud connector platforms and automation tools serve modern SaaS APIs well, with thousands of connectors for tools like Slack, Salesforce, and GitHub. Some manufacturing vendors have shipped single-vendor MCP servers that expose their own platform data only, not the systems around them.
PipeGears serves the systems all of them skip: SQL databases behind firewalls, EDI over SFTP, SOAP/XML services, MQTT brokers, and on-prem systems that have never seen the internet. One gateway. Multi-system. Governed. If your AI agent needs to talk to a legacy manufacturing system, we are the only MCP server that does it with production-grade governance.
Your AI Agent Needs Access to the Factory Floor
Get your MCP endpoint in minutes. Connect your first data source today.