imbue-mngr-kanpan 0.2.0

All-seeing agent tracker

Kanpan

All-seeing agent tracker. The name combines Sino-Japanese 看 (kan, "to look", as in 看板 kanban) and Greek πᾶν (pan, "all") -- a unified view that aggregates state from all sources (mngr agent lifecycle, git branches, GitHub PRs and CI) into a single board.

Launch with mngr kanpan. Requires the gh CLI to be installed and authenticated.

Filtering

Filter which agents appear on the board using CEL expressions:

# Show only agents for a specific project
mngr kanpan --project mngr

# Show only running agents
mngr kanpan --include 'state == "RUNNING"'

# Exclude done agents
mngr kanpan --exclude 'state == "DONE"'

--include and --exclude accept arbitrary CEL expressions (repeatable). --project is a convenience shorthand that translates to an include filter on labels.project. Multiple --project flags are OR'd together.

When any filter is active, the header displays a [filtered] indicator.

Custom commands

Add to your mngr settings file (e.g. .mngr/settings.toml):

[plugins.kanpan.commands.c]
name = "connect"
command = "mngr connect $MNGR_AGENT_NAME"

[plugins.kanpan.commands.l]
name = "events"
command = "mngr events $MNGR_AGENT_NAME"
refresh_afterwards = true

Each entry defines a keybinding (the table key, e.g. c) that appears in the status bar and runs with the MNGR_AGENT_NAME environment variable set to the focused agent's name. Custom commands override builtins when they share the same key. Set enabled = false to disable a builtin.

By default, custom commands run immediately on the focused agent. Set markable = true to make a command use dired-style batch marking instead: pressing the key marks agents, then x executes all marks at once.

[plugins.kanpan.commands.s]
name = "stop"
command = "mngr stop $MNGR_AGENT_NAME"
markable = true
refresh_afterwards = true

Custom columns

Add extra columns to the board that display per-agent data. The source field selects where the column reads from:

• "labels" (default) -- reads agent.labels[key], where key is the column's config key.
• "agent" -- reads from AgentDetails.plugin, populated by agent_field_generators via AgentInterface. Works for both local and remote agents. Requires plugin_name and field.

Values can be colored by mapping specific strings to urwid color names.

# Label-backed column (source = "labels" is the default)
[plugins.kanpan.columns.blocked]
header = "BLOCKED"
[plugins.kanpan.columns.blocked.colors]
unblocked = "light green"
blocked = "light red"

# Plugin data column: reads from AgentDetails.plugin (populated by agent_field_generators)
[plugins.kanpan.columns.waiting]
header = "WAIT"
source = "agent"
plugin_name = "claude"
field = "waiting_reason"
[plugins.kanpan.columns.waiting.colors]
PERMISSIONS = "light red"
END_OF_TURN = "light green"

By default, custom columns appear after the built-in columns (before LINK). To control the order of all columns, set column_order:

[plugins.kanpan]
column_order = ["name", "state", "custom_blocked", "git", "pr", "ci", "link"]

Built-in column names are: name, state, git, pr, ci, link. Custom columns use custom_<key> (e.g. custom_blocked for a column defined under [plugins.kanpan.columns.blocked]). Columns not listed in column_order are omitted.

When no label or plugin data is present for an agent, the column shows an empty cell.

Refresh behavior

Kanpan uses two refresh strategies:

• Full refresh (manual 'r' key, periodic 10-minute timer): fetches both agent state and GitHub PR data. Only one can be in flight at a time -- pressing 'r' while a refresh is running is ignored.
• Agent-only refresh (after push, delete, custom commands): fetches agent state without hitting the GitHub API. PR data is carried forward from the previous snapshot.

Both are configurable:

[plugins.kanpan]
# Seconds between periodic full refreshes (default 10 minutes)
refresh_interval_seconds = 600.0
# Seconds before retrying after a failed full refresh
retry_cooldown_seconds = 60.0

Refresh hooks

Run shell commands before and/or after each full refresh. Each hook runs once per agent, in parallel across all agents. Hook failures are reported as board errors but do not block the refresh.

[plugins.kanpan.on_before_refresh.notify]
name = "Pre-refresh notify"
command = "echo Refreshing $MNGR_AGENT_NAME"

[plugins.kanpan.on_after_refresh.sync]
name = "Post-refresh sync"
command = "my-sync-script"

Before-hooks run against the previous snapshot's entries (skipped on the first refresh). After-hooks run against the new snapshot's entries. Set enabled = false to disable a hook without removing it.

Each hook command receives the following environment variables:

Variable	Description
MNGR_AGENT_NAME	Agent name
MNGR_AGENT_BRANCH	Git branch (empty if none)
MNGR_AGENT_STATE	Agent lifecycle state (e.g. RUNNING, DONE)
MNGR_AGENT_PROVIDER	Provider instance name
MNGR_AGENT_PR_NUMBER	PR number (empty if no PR)
MNGR_AGENT_PR_URL	PR URL (empty if no PR)
MNGR_AGENT_PR_STATE	PR state such as OPEN, MERGED, CLOSED (empty if no PR)