Find out why overnight invoice output stopped. Fix it if the action is safe.
IBM i tools for any MCP client
Connect your AI to IBM i.
Use iSeries MCP from Claude Desktop, Pi, or another compatible AI client. Your agent can investigate jobs, logs, messages, Db2, IFS, spooled output, and 5250 sessions while access and approvals stay under your control.
- Works from the AI client you choose
- IBM i tools delivered through MCP
- Exact access and approval controls
How can I help you today?
I’ll open an approved 5250 session and inspect the native IBM i output queue.
ibmi_terminal_opentarget_alias: production · identity_alias: operationsTLS-only TN5250 session opened. Native command line is ready.
ibmi_terminal_fillrow: 22 · column: 8 · value: WRKOUTQWRKOUTQ OUTQ(QGPL/INVOICE)
ibmi_terminal_keykey: ENTER · expected_sequence: 1The batch succeeded. Delivery stopped because the approved INVOICE output queue is held.
ibmi_terminal_fill → ibmi_terminal_keyRLSOUTQ OUTQ(QGPL/INVOICE) · ENTERCommand completed. Refreshed native screen reportsStatus: RLS.
Invoice output is moving again. IBM i accepted RLSOUTQ after approval, and the refreshed 5250 screen shows the queue released.
iseries-mcp 6 terminal tools loaded
Find out why overnight invoice output stopped. Fix it if the action is safe.
I’ll open an approved 5250 session and inspect the native IBM i output queue.
production · operations
row 22 · column 8
ENTER · expected sequence 1
The batch succeeded. Delivery stopped because the INVOICE output queue is held.
RLSOUTQ OUTQ(QGPL/INVOICE) · ENTER
Invoice output is moving again. IBM i accepted RLSOUTQ after approval. The refreshed 5250 screen shows the queue released.
Submit command "RLSOUTQ OUTQ(QGPL/INVOICE)" with ENTER?
↑/↓ navigate · enter select · esc cancelFind out why overnight invoice output stopped. Fix it if the action is safe.
I’ll inspect the approved 5250 session before I change anything.
production · operations · sequence 1
WRKOUTQ OUTQ(QGPL/INVOICE)
ENTER · Work with Output Queue · status HLD · 18 files
The INVOICE output queue is held. The writer stopped with 18 spooled files waiting.
$ RLSOUTQ OUTQ(QGPL/INVOICE)Press enter to confirm or esc to cancelRLSOUTQ OUTQ(QGPL/INVOICE) · ENTER · status RLS
Invoice output is moving again. IBM i accepted RLSOUTQ after approval. The queue now reports RLS.
One server, many AI clients
Keep one IBM i control point. Change the AI when you want.
Claude Desktop, Pi, Codex, and other MCP-compatible clients can use the same controlled server. Each client gets only the IBM i tools and identities assigned to its authenticated principal.
Bring your AI
Use the agent and model that fit the work.
iSeries MCP speaks the open Model Context Protocol. Pi has a first-party hosted extension. Other compatible AI clients can connect through the standard MCP transport without moving IBM i credentials into prompts.
- ClaudeMCP-compatible
- OpenAIMCP-compatible
- GeminiMCP-compatible
- CopilotMCP-compatible clients
- PiFirst-party extension
- Any MCP clientOpen protocol
Provider support depends on the MCP capabilities of the selected client. The logo rail shows protocol compatibility, not a claim of a separate native integration for every provider.
One controlled path to IBM i
The server sits between every AI request and IBM i.
The AI client asks for work through MCP. iSeries MCP owns the private IBM i connection, decides what the authenticated principal can do, requests approval when needed, and emits redacted audit context.
Authenticate the principal
Resolve the approved target and IBM i identity. Keep hosts, profiles, passwords, and TLS settings on the server.
Expose only granted capabilities
Limit tools, jobs, queues, paths, rates, concurrent work, and sessions before a request reaches IBM i.
Run, confirm, or stop
Allow low-risk reads in scope. Require exact, one-use approval for sensitive changes. Reject mismatched arguments.
Record a safe operation trail
Project actor, capability, risk, approval, outcome, and cancellation events without credentials or raw secret values.
RLSOUTQ OUTQ(QGPL/INVOICE)See controlled AI access to IBM i
Bring one IBM i workflow. See how your AI can handle it.
We will show the MCP connection, server-level access control, approvals, and audit context around a realistic IBM i task.
Case studies are in preparation. Request a demo to discuss your IBM i workflow now.
