MCP Agent Tools

OL TermIQ exposes Model Context Protocol (MCP) tools for AI agents and automation. Recorded-data tools use /api/v1/mcp. Live and mock terminal tools use /api/v1/mcp/terminal.

Recorded data (read-only, 8 tools)

• list_vocabulary: available filter dimensions and report categories
• describe_screen: catalog screen details and outgoing transitions. A matched map is reported as map=
• describe_screen_maps: list uploaded BMS, MFS, and DDS map sections, or inspect one section's fields, labels, tables, PF keys, and MFS messages
• search_terminal_events and get_terminal_events: find and inspect keystroke-level events
• list_terminal_sessions: browse recorded sessions
• run_terminal_report: execute ad-hoc reports
• get_screen_transitions: recorded actions leaving a catalog screen, including unpublished New screens

Terminal server (7 tools)

Every call sends x-ol-termiq-key, which names the tenant, plus a bearer token from the host sign-in. On first use the client opens a browser page: choose the workspace, host connection, and sign-on macro, or enter the host user and password. Host credentials are kept encrypted for the sign-in lifetime (7 days by default) so a refresh or a restart can sign on again.

• list_connections: host connections of the chosen workspace
• connect_session: open the terminal on the host
• get_screen: read the current screen with catalog-resolved identity
• execute_action: press an AID and optionally set field values by name. It is not replayed if the host ended the session
• disconnect_session: close the terminal. An open recording is saved
• start_trail_recording and stop_trail_recording: record the signed-in live session. The sign-on is not included. A mock session cannot be recorded. A TermIQ started with OL_NO_DT refuses both

Trail playback

A mock connection plays its trail in order. A host key (Enter, a PF key, and similar) moves to the next screen. Local keys such as Tab stay on the current screen. Playback writes no terminal events. The last host key ends the playback, and get_screen starts it again.

Configure terminal connections, trails, and sign-on macros under Workspace, then Configuration.