Following its Beta introduction last quarter, the Transfer CFT MCP (Model Context Protocol) Server is now generally available. Customers can connect their own AI assistants and automation tools to Transfer CFT through a standardized interface, building on the AI technologies they already use, allowing for AI-augmented operations.

The MCP Server now extends beyond Windows and Unix to all Unix platforms, HP NonStop, IBM i, and z/OS, so organizations can adopt AI-assisted monitoring, troubleshooting, reporting, and administration across their entire estate, including mainframe and midrange environments.

For governance and auditability, Transfer CFT now tags every object created or updated through the MCP/AI server with ORIGIN=MCP, which keeps AI-driven operations transparent and traceable.

This release gives teams more precise control over how transfers are handled.

New routing step options were added on all routing step types, including archive, extract, routing, and notification steps: the ALLOWFAILURE and WHEN parameters. Use ALLOWFAILURE to decide whether a step failure stops route processing, and WHEN to decide when a step runs. Routes can now adapt to the outcome of previous steps, reducing the need for custom workarounds.

Transfer CFT supports now case-insensitive folder monitoring: Filename filters in Folder Monitoring (CFTFOLDER) can now ignore case, using the new IWILDMAT, IEREGEX, and ISTRJCMP FILTERTYPE values. This is especially useful in mixed environments where file naming conventions are inconsistent.

The maximum number of parallel connections is now 2,000 on Unix, Windows, and mainframe systems, doubling the previous limit of 1,000.  For workloads that exceeded that limit, this reduces retries and improves transfer throughput over TCP and pTCP, helping organizations handle growing volumes with confidence, so large batches complete faster and peak-period workloads are absorbed more smoothly.

Enabling the higher limit is simple: administrators set MAXCNX=2000 in CFTNET to open up to 2,000 simultaneous connections. The increase applies across all protocols supported PeSIT, Odette, and SFTP, so it benefits the full range of partner connections. It works for single-node setups as well as multi-node deployments, where each node supports up to 2,000 connections. For large-scale distribution, a Transfer CFT client can now broadcast files to up to 4,000 Transfer CFT servers.

Transfer CFT now lets organizations manage internal users and groups (CFTUSER and CFTGROUP) through all its interfaces: the UI, REST APIs, and CLI.

A configurable password policy is enforced on user creation and password change (am.internal.user.password_policy.* parameters) , so teams can apply their own internal security standards, including: Password length and complexity, Lockout after unsuccessful logon attempts, Password expiration, Password history.

This makes it easier to manage users on externally managed infrastructure while meeting corporate security and compliance requirements. It is available on Unix and Windows.

The quick search bar is now available from any page of the Web UI, including logs, security, flow definition, and unified configuration pages. Teams can find what they need faster, which reduces administrative effort and speeds up day-to-day operations.

The search supports free-text filtering across key fields, using a “contains” match to quickly locate relevant entries. You can also enter multiple search values to refine your results and make filtering more precise.