Dashboard views
The dashboard is a small set of routes served from http://localhost:6006. This page walks through each one — what it shows and what you use it for. For the model behind the agent / module / flow views, see Concepts.
Kanban board
The board is the landing page (/). It lays your tasks out in columns, one
column per status in the bound flow.
Screenshot to add: the Kanban board showing tasks in flow-status columns.
(save as static/img/dashboard/kanban.png, then embed here)
- Shard navigation — tasks are stored in shards of ten. When a project has more than one shard, a shard switcher appears above the board so you can page through them.
- Aggregate stats — a strip of summary counts (totals and per-status tallies) sits above the columns.
- Flow-status columns — the columns are derived from the statuses of the bound flow, so a custom flow with custom states shows its own columns. Click any card to open its detail panel.
- Settings popover — the ⚙ button opens notification settings: toggle Sound and Mute when tab focused for the desktop / sound alerts that fire when an agent finishes or needs permission.
When the activity engine is enabled, the board also carries the Agent
Activity and Recent Activity tabs (see Activity feed
below). If the engine is off, a chip explains how to enable it in
taskflow.config.json.
Task detail
Open any card, or visit /task/<id>, for the full record of one task.
Screenshot to add: the task detail page with metadata, status history, reviews, and the lifecycle map.
(save as static/img/dashboard/task-detail.png, then embed here)
The page gathers everything about the task in one place:
- Metadata — title, type, priority, tags, and the assigned flow.
- Status history — the trail of status transitions the task has moved through.
- Reviews — the recorded AI and human reviews, with their verdicts.
- Incidents — any production incidents logged against the task.
- Implementation details — the implementation notes and artifacts captured as the task progressed.
- Lifecycle position — a map of where the task sits in its flow, with
suggested next agents (each chip copies the agent's slash command). While the
task is still
readyyou can reassign it to a different flow; once work starts, the flow locks.
Agents browser
The agents browser (/agent, or /agent/<id>) lists every composed
agent in the sidebar with its title and description.
Screenshot to add: an agent detail page with its composition map of modules.
(save as static/img/dashboard/agents.png, then embed here)
Selecting an agent shows its composition map — the ordered sequence of modules (sections, includes, MCP servers, hooks, skills, bundles, and behavior nodes) that compose into the agent's prompt. Clicking a module node opens an info modal in place. The header shows which slash command the agent installs as, and Install / Uninstall buttons apply the agent's prompt and its modules' artifacts to your project. See the shipped agents for the defaults you start with.
Modules browser
The modules browser (/module, or /module/<id>) is the full
module registry, grouped in the sidebar by custom,
shared, integrations, and per-agent modules.
Screenshot to add: a module detail page showing its kind-specific panel and composition map.
(save as static/img/dashboard/modules.png, then embed here)
Each module's detail page shows its kind badge, source, and description, plus kind-specific panels — section body, include reference and preview, MCP config JSON, hook event/matcher/command, skill content, bundle contents, or a behavior node's fields. A referenced by panel lists the agents that compose the module, and a composition map ties it all together. Installable kinds (MCP server, hook, skill, bundle) offer Install / Uninstall actions. The shipped modules page documents the defaults.
Flow / project editor
The flow editor (/project, or /project/<id>) is the top layer: each named
flow, with its agents, lifecycle graph, statuses, entry
points, and install list.
Screenshot to add: the flow editor with the lifecycle graph and the sidebar of agents and installed modules.
(save as static/img/dashboard/project.png, then embed here)
- Flows — every flow is listed in the sidebar (the shipped default is badged). Custom flows can be deleted, set as the binding default, or duplicated as a starting point for a new one.
- Agents — the flow's agent set, with its entry agents (start points) marked ★.
- Flow graph — a map of the lifecycle: nodes are agents, edges are the handovers between statuses. In edit flow mode you can drag nodes, wire edges (picking the trigger status for each), and manage custom states that alias onto the canonical task statuses.
- Install list — the modules installed in this flow, with Install this flow / Uninstall to apply or remove the flow's MCP servers, hooks, and skills.
New flows are created from /project/new. The shipped default ejects to a
custom copy the first time you save edits to it, and can be reverted to the
shipped version.
Activity feed
When the activity engine is enabled, the board surfaces a live feed of what Claude is doing.
Screenshot to add: the activity feed showing the live Claude status and an event timeline.
(save as static/img/dashboard/activity.png, then embed here)
- Live Claude status — a status chip reflects whether the agent is active, idle, awaiting permission, or done. The same signal drives the page-title glyph, sound, and desktop notifications you toggle in the board's settings popover.
- Event timeline — a stream of activity events (tool use, status changes) rendered as they arrive, filtered by the configured verbosity.
Everything here arrives over the same server-sent-events stream the rest of
the dashboard uses, so the feed is live without a refresh. Enabling the engine
is a one-line change in taskflow.config.json; when it
is off, the dashboard falls back to a static recent-activity timeline.