Make legacy and custom systems AI-ready in hours with Frends Enterprise MCP

Let your LLM of choice — Claude, OpenAI, or your own — talk to your systems with governance intact. No re-engineering. No rebuilding. No extra cost.

Enterprise MCP Hero
Enterprise MCP Hero

Enterprise systems weren't built for AI. Until now.

Enterprise AI stalls because custom and legacy systems can't interact with AI agents. With Frends Enterprise MCP, you add a governed gateway in front of any system, making it accessible to your AI of choice.

Legacy and custom systems had no MCP interface

ERP platforms, finance systems and in-house applications couldn't be discovered or called by an AI agent. Now you can remove your legacy systems' blindfolds and make them MCP native.

Re-engineering systems or custom coding isn't an option

Most organizations can't justify rebuilding years of working infrastructure just to create an AI interface for each system. Frends Enterprise MCP bridges this gap with a solution that requires no code and is hybrid by design.

Governance and control don't have to be a blocker

AI platforms require separate security models with limited visibility. IT has now a unified view of what agents are doing and in which systems, and can manage AI access from the same iPaaS dashboard already use.

The Solution: Frends Enterprise MCP turns any existing system into a governed, AI-callable tool

Make any custom or legacy system immediately discoverable and callable by your AI of choice. With every action logged, auditable and governed by the same API policies already securing your REST APIs. 

Key capabilities

Let any AI talk to any system

Add an MCP Trigger to a Frends Process and any custom or legacy system becomes a governed AI tool, reachable by Claude, OpenAI or your own agent in just a few hours.

Autonomous AI reasoning with scoped tool access

The Frends AI Connector runs step-by-step reasoning loops where each AI step uses only the tools it is explicitly authorized to reach. Complex workflows, broken into manageable, governed AI steps.

Full governance, visibility and audit trail

Every AI action is logged as a standard process execution. The full reasoning chain is visualized in process workflows, readable by IT, compliance and business stakeholders alike, without black boxes.

See Frends Enterprise MCP in action

Take the interactive demo to see how the Frends Enterprise MCP works, from exposing a legacy system as an AI tool to watching an autonomous agent reason step by step across your systems.

What this means for your business

Here's a preview of what Frends Enterprise MCP enables:

Modernize legacy systems with AI access in hours-2

There is no need to wait for a major re-architecture project before AI can start working. Enterprises that previously faced months of custom integration work are up and running in hours.

Connect your AI of choice to any custom or legacy system-2

From Claude to OpenAI, GPT or your own agent, any MCP-compatible AI can discover and call your systems through Frends. Without vendor lock-in. 

Keep what you’ve built, and make it AI-ready-2

 Custom applications, legacy ERPs, in-house APIs — none of them need to change. Frends wraps them as governed AI tools exactly as they are. Your existing investment is preserved and made AI-ready at the same time.

Deploy AI with confidence in regulated industries.-2

 Every AI interaction is logged, BPMN-visualized and governed by existing API policies. Frends Enterprise MCP is EU AI Act-ready and GDPR-native, built by a European vendor not subject to the CLOUD Act. Compliance teams get human-readable audit trails without any new tooling.

Run your AI tools anywhere_ cloud, on-premises or boht.-2

Frends Agents run as MCP Servers on-premises, in the cloud, or across a hybrid architecture. Sensitive systems can stay fully on-premises, including with local AI models, keeping data within sovereign boundaries.

Modernize legacy systems with AI access in hours-2 Connect your AI of choice to any custom or legacy system-2 Keep what you’ve built, and make it AI-ready-2 Deploy AI with confidence in regulated industries.-2 Run your AI tools anywhere_ cloud, on-premises or boht.-2

Use cases

See it working in the real world

Custom & legacy systems
Custom & legacy systems
Finance (invoice handling)
Finance (credit risk)
Customer support

AI reaches the full system estate from day one

A large enterprise wants their AI to interact with multiple in-house systems — ERP, finance platforms and custom-built applications — without rebuilding any of them. Frends wraps each system as a governed MCP tool in hours. The AI of their choice can now reach the full system estate from day one.

Custom & legacy systems

Accounts Payable exception handling 85% faster

An AI triage agent checks incoming invoices against ERP and finance systems via MCP tools, applies tolerance policies and auto-resolves 60–80% of exceptions without human touch. Complex cases are escalated with an AI-written brief, cutting review time from 15 minutes to 3–5 minutes. Overall AP exception handling time drops 70–85%.

Finance (invoice handling)

Credit risk pre-screening in seconds

An AI agent retrieves legal entity data and internal loan exposure from separate legacy and custom systems in parallel, calculates aggregate exposure, flags risk thresholds (critical, high exposure, review required) and routes the case automatically. What previously required a specialist and multiple system lookups is done in seconds.

Finance (credit risk)

End-to-end IT incident resolution

An AI agent receives an incident ticket, calls diagnostic MCP tools to check system health, identifies the issue, restarts the failing service, verifies resolution and closes the ticket, all without waking anyone up. The on-call engineer finds a summary report in the morning.

Finance (credit risk)-1

Webinar: Introducing Frends Enterpriser MCP

Join us live to see Frends Enterprise MCP in action, from the MCP gateway pattern to live demos of autonomous AI agents working across real enterprise systems. June 23, 14:00 CEST.

FAQ

Frends Enterprise MCP is a feature set in Frends 6.3 that enables any Frends Process to be exposed as a standardised AI tool via the Model Context Protocol, an open standard for AI-to-application communication. It consists of two components: the MCP Trigger, which turns any process into a governed AI-callable tool, and the Frends AI Connector, which enables autonomous AI reasoning loops with scoped tool access inside Frends workflows.
You add an MCP Trigger to a Frends Process that sits in front of your existing system — legacy ERP, custom application, internal API, or any other. That process becomes a governed gateway: your AI of choice discovers it via the MCP tools/list command and calls it via tools/call. The underlying system receives a standard API call, it never needs to know an AI was involved, and you never need to change a line of its code.

Any MCP-compatible AI client can connect to Frends as an MCP Server, including Claude, ChatGPT, custom agents, and the Frends AI Connector itself. For AI model usage inside the Intelligent AI Connector, Frends supports Frends-hosted models, your own Azure AI Inference service, on-premises Ollama installations, and any custom AI service with a compatible API.

Yes. Each Frends Agent acts as an independent MCP Server and can run fully on-premises. This includes support for local AI models via Ollama, keeping all data and AI interactions within your own infrastructure — with no dependency on Frends cloud services. This makes Frends Enterprise MCP suitable for air-gapped environments and organisations with strict data sovereignty requirements.

MCP endpoints are secured using Frends' existing API Policies framework, the same framework that governs all REST APIs on the platform. Each AI agent authenticates via OAuth 2.0 or JWT tokens. Access is permission-based: an AI agent only sees the tools it is explicitly authorised to access. Unauthorised tools are invisible to the caller. Every interaction is logged with agent identity, parameters passed, output returned, and execution duration.
Every MCP tool call is logged as a standard Frends process execution and visualised in the BPMN canvas, including the AI's full reasoning chain when using the Intelligent AI Connector v2. Logs include timestamp, AI agent identity, tool called, input parameters, output, and execution duration. Logs are immutable, retained according to your tier's data retention policy, and exportable via the Frends Management API. This directly supports EU AI Act transparency requirements for high-risk AI.
No. Adding an MCP Trigger to a Frends Process does not change how that process runs for any existing callers. The Frends AI Connector is fully backward compatible, existing AI tasks continue to function as before until you explicitly enable Agentic mode. Your current integrations are unaffected.
The MCP Trigger carries no additional licensing fee and is included in Business and Enterprise tiers. Each MCP tool call counts as one process execution against your existing monthly limit. The Frends AI Connector is also included in Business and Enterprise tiers executions consume Frends Credits (€0.01/credit), only when using Frends-hosted AI models. If you bring your own AI — Azure OpenAI, Ollama, or a custom model — you pay zero Frends Credits for AI model usage.

Governance is a core principle of the Frends AI implementation. It is ensured through: 

  • Full audit trails: Every AI interaction, including the prompt, output and reasoning, is logged for complete traceability. 
  • Transparent reasoning: The AI's "chain-of-thought" is captured, so you can see exactly how it reached a conclusion. 
  • Cost monitoring: Real-time dashboards track token usage and estimate costs to prevent budget overruns. 
  • Human-in-the-loop: You can configure checkpoints where a human must approve an AI decision before the process continues. 

Ready to make your custom and legacy sustems AI-ready?

See how Frends Enterprise MCP connects your AI of choice to your custom and legacy systems, in a live, personalised demo with our team.