IBM Mainframe (z/OS, CICS, IMS)
Connect AI agents and workflows to the IBM mainframe: CICS and IMS transactions, Db2 for z/OS, JCL batch jobs and datasets, through z/OS Connect.
Book a demoIBM Mainframe (z/OS, CICS, IMS) for AI agents and workflows.
engini lets your AI agents and workflows work with the IBM mainframe that runs your core banking, insurance or logistics: call CICS and IMS transactions through z/OS Connect, query and update Db2 for z/OS, submit JCL batch jobs and read their output, and read and write datasets. The programs stay exactly as they are, and the On-Prem Agent connects out to engini with no inbound ports. Connect from Claude, ChatGPT or Cursor over MCP, from your code, or build in Workflow Studio.
Cursor, your code
audit log
in engini
What your agent can do in IBM Mainframe (z/OS, CICS, IMS). Every tool is one you allow or leave out.
The IBM Mainframe (z/OS, CICS, IMS) tools in engini, grouped by what they change. Read tools are safe to hand out widely. Write and delete tools go only to the agents that need them.
Read
Run a SQL query on Db2 for z/OS.
Follow a batch job and read its spool output.
Read a sequential dataset or a PDS member.
List datasets that match a pattern.
Write
Call any REST API your team exposed through z/OS Connect, backed by a CICS program, an IMS transaction or Db2.
Run a CICS program with its COMMAREA or channel, through z/OS Connect.
Run an IMS transaction, through z/OS Connect.
Write to Db2 for z/OS tables in a transaction.
Submit JCL, from a dataset or inline, through z/OSMF.
Write a sequential dataset or a PDS member.
Triggers
A program put a message on an IBM MQ queue engini watches.
A batch job you follow finished, with its return code.
Rows added to a Db2 for z/OS table.
A trigger starts a workflow, and a workflow can include an AI agent step. In Code, a trigger delivers the event to your own endpoint.
Connect your AI agent to IBM Mainframe (z/OS, CICS, IMS). Four ways in, one set of permissions.
Claude, ChatGPT, Cursor
Create an MCP server in engini with your IBM Mainframe (z/OS, CICS, IMS) connection and only the tools you choose. Add its address to your client and sign in through engini. There is no key to copy.
MCP client setup →Python and TypeScript
Hand your agent a toolset of IBM Mainframe (z/OS, CICS, IMS) tools. The SDK gives OpenAI and Anthropic models the tools in the format they expect and runs the ones they pick.
SDK guide →For coding agents
The engini CLI connects IBM Mainframe (z/OS, CICS, IMS) and runs its tools from a terminal, with machine-readable output, so an AI coding agent can drive it without getting stuck.
CLI guide →Any language
Every IBM Mainframe (z/OS, CICS, IMS) tool is an endpoint. Call it with exact inputs from a scheduled job or a sync, with no model in the loop.
API reference →.cursor/mcp.json. Claude Code: claude mcp add.// .cursor/mcp.json: Cursor, connected to IBM Mainframe (z/OS, CICS, IMS) through engini
{
"mcpServers": {
"engini-ibm-mainframe": {
"url": "<your engini MCP server address>"
}
}
}
Create the MCP server in engini with your IBM Mainframe (z/OS, CICS, IMS) connection and the tools this agent may use, then paste its address. The client signs in through engini on first use. Claude, VS Code and other clients take the same address: see the MCP client setup guide.
Each request runs through the IBM Mainframe (z/OS, CICS, IMS) tools in the agent's toolset or MCP server. If a tool is not allowed, the agent cannot call it.
What teams build on IBM Mainframe (z/OS, CICS, IMS). Agents and workflows, across systems.
Agents that reach the core
A customer service agent reads balances and transactions from the core banking system on the mainframe and answers in Teams. Its toolset holds read tools only.
Claims from the portal into the policy system
A claim from the web portal or an email is checked against the policy through an IMS transaction, then opened in the claims system.
Batch cycles people can see
When the nightly batch ends, its return codes and exception reports go to Slack and ServiceNow.
Mainframe data in the cloud
Db2 for z/OS tables land in Snowflake on a schedule, and the mainframe keeps doing what it does.
Every integration runs both ways, so any of these flows can also run in reverse. These are examples, not the limit: any flow between IBM Mainframe (z/OS, CICS, IMS) and another system engini connects is available.
You decide what each agent may do in IBM Mainframe (z/OS, CICS, IMS). engini enforces it on every call.
Only the tools you allow
A toolset lists the IBM Mainframe (z/OS, CICS, IMS) tools an agent may call and the connection each one uses. Give one agent read tools only and another the write it needs. engini refuses any call outside the toolset.
Scoped when you set them up
An MCP server exposes only the connections and tools chosen when it is created, and engini shows who uses each server.
Whose permissions apply
Connect a person's own IBM Mainframe (z/OS, CICS, IMS) account and engini acts as that person, with their IBM Mainframe (z/OS, CICS, IMS) permissions. Use a shared integration account and it acts as that account. Every run is recorded under the user who started it.
Kept out of prompts
Your agent never needs your credentials: engini stores them and adds them when it runs each call, so keys stay out of prompts and tool inputs.
Every call on record
Every call lands in History and the audit log, which you can export or stream to your SIEM.
Test before production
Workspaces keep teams apart, and development, staging and production environments keep a test agent away from live data.
The mainframe behind your firewall.
If the mainframe runs in your own data center, install the engini On-Prem Agent on a server inside your network. It opens every connection itself: out to engini over HTTPS on port 443, and to the mainframe while it runs a task. You open no inbound ports or firewall rules. Calls through it run on the same engine as MCP, the API and workflows, and each one is recorded in History.
On-Prem Agent guide →Connect IBM Mainframe (z/OS, CICS, IMS) in three steps. Then decide what each agent may do.
Sign in to IBM Mainframe (z/OS, CICS, IMS)
Install the engini On-Prem Agent on a server that can reach z/OS Connect and z/OSMF, create a RACF user or certificate with access only to the APIs, tables, datasets and job classes the workflows need, then add a mainframe connection with those addresses.
Choose the tools
Add the connection to an MCP server or a toolset with only the tools each agent needs.
Build or ask
Use the connection as a step in a workflow, or ask your agent to work in IBM Mainframe (z/OS, CICS, IMS) from Claude, ChatGPT, Cursor or your own code.
IBM Mainframe (z/OS, CICS, IMS) and engini. Straight answers.
Can an AI agent write to IBM Mainframe (z/OS, CICS, IMS), or only read?
Both. engini has tools to call z/OS Connect APIs, run CICS programs and IMS transactions, write Db2 for z/OS tables, submit batch jobs and write datasets. Whether a given agent may use them is up to you: give it a toolset or an MCP server with read tools only, and it cannot call the rest.
Do we need to change our COBOL programs?
No. z/OS Connect exposes the CICS programs and IMS transactions you already have as REST APIs, so the programs and their rules stay as they are.
What if the tool I need is not listed?
Expose it through z/OS Connect and call it with Call z/OS Connect API, or run a batch job with Submit Job. Give them only to the agents that need them.
What do we need on the mainframe side?
IBM z/OS Connect for CICS and IMS, z/OSMF for jobs and datasets, and network access from the On-Prem Agent. Most z/OS shops already run both.
Does the agent act as me in IBM Mainframe (z/OS, CICS, IMS)?
It acts as the RACF user or certificate behind the connection, with exactly the access RACF gives it. Every run is recorded under the person who started it.
Can a change in IBM Mainframe (z/OS, CICS, IMS) start work?
Yes. A message on an IBM MQ queue engini watches starts a workflow as it arrives, and scheduled triggers follow batch jobs and check Db2 tables for new rows. A workflow can include an AI agent step.