ARCAD MCP Server New Feature!

Give your AI agents your application context

Connect your AI tools to your IBM i applications with ARCAD MCP Server and get an AI that truly understands your system.

AI assistants code well, but don’t know your system.

LLMs master language, but have no idea what your programs do, how your data is structured, or how your applications interact.

“What AI assistants lack is the knowledge of your system. They all hit the same wall: contextualization.”

— Marc Dallas, R&D Director, ARCAD Software

Exposing the code: too risky

Scanning all source code raises cost, dependency, and confidentiality issues.

Without context, AI hallucinates

Without knowledge of your architecture, assistants generate generic or incorrect code.

Application knowledge

Legacy experts are retiring; new developers query applications in natural language.

Legacy code resists AI

40 years of evolution (RPG/3, RPG/4, free format…) produce heterogeneous coding styles that are difficult to analyze.

40 years of deterministic repository

70+ MCP tools available

Multi-LLM Compatible on-premise & cloud

Deterministic context for your AI agents

Expose the ARCAD repository via the MCP protocol: precise, factual, and secure data.

1

ARCAD Repository

Your own metadata

ARCAD has been analyzing your source code, production data, and dynamic code for 40 years using deterministic algorithms. The result: a complete, reliable, and always up-to-date repository: components, fields, procedures, literals, dependencies.

2

MCP Protocol

Secure service exposure

The MCP server exposes 70+ tools (resources, actions, methodological prompts) from the ARCAD repository. Only the information needed for the current task is transmitted to the agent, not the entire codebase.

3

AI Agent

The assistant chains calls autonomously

The AI (Copilot, Bob, Claude, or your on-premise LLM) queries MCP tools in natural language, launches multiple services in parallel, consolidates results, and delivers a concise and accurate response: diagram, impact analysis, contextualized code.

4

Developer

You stay in control

The developer remains in the decision and validation loop. AI accelerates, structures, and documents. The human expert validates and arbitrates. The best of both worlds.

A hybrid approach for maximum accuracy

Probabilistic AI + Deterministic Analysis = ARCAD MCP Server

Certified context, not estimated

Data injected into the AI comes from deterministic algorithms, not LLM inference. Zero hallucination.

Asset security

Only the elements needed for each task are exposed. Your complete source code never leaves your environment.

On-premise compatible

Works with on-premise LLMs. Since contextualization compensates for raw model power, even a local SLM produces relevant results.

Multi-IDE, multi-agent

Native integration with VS Code Copilot, Bob (IBM i), and any agentic orchestrator (Make, N8N…). A single standard protocol.

Multi-platform coverage

IBM i today, z/OS tomorrow. DISCOVER acts as a transparent context gateway regardless of the analyzed source.

Quarterly updates

New batches of MCP tools are released every 3 months, progressively covering the entire ARCAD suite (DROPS, DISCOVER, DOT…).

Access ARCAD MCP Server from your environments

VS Code + GitHub Copilot
Bob (IBM i)
Claude
Make / N8N
On-premise LLM
IBM i (AS/400)ARCAD for DevOps
DISCOVER
DROPS for i
DROPS for z/OS (coming soon)

Get your IBM i ready for the agentic era

ARCAD MCP Server is available today with 70+ tools.

Personalized Demo

Talk to our experts