MCP Tools¶

Narnia exposes sixteen MCP tools: eight for working with your local Copilot CLI session history, and eight for managing Narnia-owned scheduled Copilot jobs. Use them from any AI assistant that supports the Model Context Protocol — including GitHub Copilot CLI, Claude Desktop, Cursor, and VS Code.

Session History Tools¶

Tool	Purpose	Key Parameters
`list_recent_sessions`	List most recently updated sessions	`limit`
`search_sessions`	Ranked session-content search	`query`, `limit`
`get_session_details`	Full metadata for one session	`sessionId`
`get_session_checkpoints`	All checkpoints for a session	`sessionId`
`get_session_turns`	Paginated conversation turns	`sessionId`, `offset`, `limit`
`get_session_workspace`	Git root and artifact file list	`sessionId`
`list_sessions_by_repository`	Exact effective remote-repository filter	`repo`
`list_sessions_by_cwd`	Exact working-directory filter	`cwd`

Looking for the web UI launcher?

The open_narnia_ui MCP tool has been replaced by the narnia-web-server skill, which provides more reliable lifecycle management with full visibility into build output.

Common Workflow¶

After a machine restart, choose the field you actually know:

list_recent_sessions — scan what was running recently
list_sessions_by_repository or list_sessions_by_cwd — filter exact metadata
search_sessions — search conversation content by topic
get_session_details — confirm the right session before resuming
get_session_checkpoints — read the last checkpoint to restore context

All session tools return JSON. Results are read-only — Narnia never modifies your session data.

Scheduled Job Tools¶

Manage Narnia-owned scheduled Copilot jobs — recurring copilot -p runs that Windows Task Scheduler executes unattended on a daily/weekly/monthly cadence. See the narnia-scheduler skill for the full create/migrate/verify workflow built on top of these tools.

Tool	Purpose	Key Parameters
`list_schedules`	All cataloged jobs joined to live task status	—
`get_schedule`	A single job's full catalog entry by id	`id`
`get_schedule_log`	Latest run log and live running state	`id`
`create_schedule`	Create a job and (by default) register its task	`name`, `prompt`, `cadenceKind`, `time`
`update_schedule`	Replace a job's definition and re-register	`id`, `name`, `prompt`
`set_schedule_enabled`	Enable/disable a job's task	`id`, `enabled`
`run_schedule_now`	Start a job's task immediately	`id`
`delete_schedule`	Remove a job's task, wrapper, and catalog entry	`id`

These tools are backed by the same service as the web UI's Schedules page — creating or editing a job through either surface is immediately visible in the other.