"
+```
+
+Treat the password as a shared read (and light write) credential once
+you've given it to an agent. Anyone holding it can query the
+organization's activity, tasks and tool usage, and add or apply labels.
+
+## Filters
+
+`query_activity`, `query_tasks`, `query_task_quality` and `query_tool_usage`
+share one set of filters:
+
+| Filter | Meaning |
+|---|---|
+| `since`, `until` | Date range (`YYYY-MM-DD`), inclusive. Defaults to the last 30 days if omitted. |
+| `project` | Substring match on the project path. |
+| `source` | One of `claude`, `codex`, `gemini`, `cowork`. |
+| `user` | One `userId`, from `query_users`. Omit for the whole organization. |
+| `group` | A `groupId`, or `__none__` for people with no group. |
+
+`query_tasks` adds `q` (free-text search), `outcome` (comma-separated
+`success`, `failure`, `unknown`) and paging with `limit` (default 50,
+maximum 200) and `offset`. `query_users` only takes `group`. It's a
+roster, not a windowed report, so it's the tool to call first to find a
+`userId` before scoping the others.
+
+## Tools
+
+| Tool | What it answers |
+|---|---|
+| `query_users` | The roster: user IDs, display names, emails, group, last-sync time, sessions, tokens and cost. |
+| `query_activity` | Usage and cost over the window, plus the same figures for the prior window for comparison. |
+| `query_tasks` | A paged, filterable list of tasks, with outcome counts for the filtered set. |
+| `query_task_quality` | Success, frustration and friction rates, outcomes over time, and the top failure signals. |
+| `query_tool_usage` | Which tools, skills and MCP servers people use, and how usage compares across sources. |
+| `list_labels` | Every [Argus Hub label](/argus-hub/tasks#argus-hub-labels) defined on this Argus Hub, with how many tasks carry it. |
+| `create_label` | Adds a new Argus Hub label. |
+| `set_task_label` | Applies or removes an Argus Hub label on one task. |
+
+A query tool with no matching data returns an empty result (an empty task
+list, an empty roster) rather than an error. `query_activity` and
+`query_task_quality` are the exception: they return an error when the
+organization has no data at all in the window.
+
+## A worked example
+
+Ask an agent something like:
+
+> Who on the team has the highest Claude Code spend this month, and does
+> their task success rate look normal?
+
+A reasonable path: call `query_users` to find people, `query_activity`
+scoped to `source: "claude"` and this month's `since`/`until` to rank
+spend, then `query_task_quality` with the same filters (or `user` set to
+the top spender) to read their success and frustration rates against the
+organization's.
+
+Every tool's response comes back as a JSON string in the result content,
+so have the agent parse it rather than expect structured fields directly
+on the MCP response.
diff --git a/docs/argus-hub/tasks.md b/docs/argus-hub/tasks.md
new file mode 100644
index 00000000..f889fbdf
--- /dev/null
+++ b/docs/argus-hub/tasks.md
@@ -0,0 +1,98 @@
+---
+description: How Argus Hub rolls up task outcomes, frustration and failure signals across the organization, and how to label tasks for your team.
+---
+
+# Argus Hub: Tasks
+
+Argus Hub's Tasks view takes the same per-session [task](/terminology#task)
+extraction described in [Tasks](/tasks), a task's description, its outcome
+and how much friction it took, and rolls it up across everyone syncing to
+the Argus Hub. Where the single-client Tasks view shows one session at a time,
+this view is for spotting patterns: which projects or people are hitting
+trouble, and what's causing it.
+
+
+
+
+
+
+
+## Filters
+
+The same filter bar as [Activity](/argus-hub/activity): date range, source,
+group. This view adds two filters of its own: a free-text search over task
+descriptions and projects, and an outcome filter (success, failure,
+unknown) you can combine with it.
+
+## Headline totals
+
+Four stat cards: **Total tasks**, **Success rate**, **Frustration rate** and
+**Interrupted rate**. Each rate is a share of the tasks with a clear
+reading. A task Argus couldn't classify doesn't count against you, and a
+card reads as unavailable (`—`) rather than 0% when nothing in the window
+has a clear reading yet.
+
+## Outcome and frustration
+
+Two breakdowns, each as a donut: **Outcome** (success, failure, unclear) and
+**Frustration** (none, moderate, high). Both come from the same free-text
+judgment a session's interpretation writes for each task. Argus Hub just
+classifies and counts it across everyone, using the same rules a single
+client uses, careful about phrases like "not completed" so a negated
+success doesn't get miscounted.
+
+## Trend over time
+
+A stacked daily bar chart of task outcomes, success, failure and unclear,
+one bar per day across the whole window, including days with no tasks.
+
+## By user, by source, by project
+
+Three ranked tables, each with task count, success rate and frustration
+rate, sorted by volume: **by user**, **by source** and **by project**. Like
+the per-user rankings on Activity, the by-user table stays hidden until the
+organization has at least three people, so a small team's tasks aren't
+attributed by name.
+
+## Signals and friction
+
+**Top signals** ranks the short failure-signal tags Argus attaches to tasks
+that failed or came out frustrating, the top ten, most common first.
+**Friction** rolls up interruptions, declined tool actions and context
+compactions the same way [Health](/metric-views#health) does for one
+client. It's measured for Claude sessions only, so a count of zero and "no
+data" show as different things, and an observed-sessions count makes clear
+how much of the window that friction reading actually covers.
+
+## Argus Hub labels
+
+Argus Hub labels are separate from the labels you set on a session in the Argus
+app itself (see [Sessions](/sessions#labeling-and-hiding-sessions)). Those
+stay local to your machine and aren't part of what
+[sync](/terminology#sync) uploads. An Argus Hub label instead lives on the Argus Hub
+itself, created by whoever runs it, and applied directly to the tasks Argus Hub
+has collected. Use them to flag or group tasks across the whole
+organization: tasks worth reviewing, tasks tied to a project or
+initiative, tasks a specific team should look at.
+
+- **Create a label.** Open **Labels** to see every label defined on this
+ Argus Hub. Add one with a name and an optional description of what it means
+ (shown when you create it, not in the list). Querying labels over
+ [MCP](/argus-hub/mcp) also returns how many tasks currently carry each one.
+- **Apply a label.** On a task in this view, open its label picker to find
+ an existing label or create one on the spot, and toggle it onto that
+ task. Applied labels show as pills on the task, both in the list and once
+ expanded.
+- **Remove a label.** Toggle it off the same way, or use **Clear** on a
+ task to remove every label from it at once.
+- **Delete a label.** Deleting a label from the Labels page removes it from
+ every task that carried it.
+
+
+
+
+
+
+
+Argus Hub labels never travel back to a client. They exist only in the Argus Hub's own
+database, for whoever has access to the Argus Hub dashboard.
diff --git a/docs/argus-hub/team.md b/docs/argus-hub/team.md
new file mode 100644
index 00000000..1d6ff07e
--- /dev/null
+++ b/docs/argus-hub/team.md
@@ -0,0 +1,54 @@
+---
+description: The per-person summary table, organizing people into groups, and each person's own activity page.
+---
+
+# Argus Hub: Team
+
+Team is the roster: every person whose Argus has ever synced to this Argus Hub,
+what they've used, and when they last synced. It's also where you organize
+people into groups so Activity, Tasks and Tools can be scoped to a team
+rather than the whole organization.
+
+
+
+
+
+
+
+## The table
+
+The table lists people in sections, one per group plus a trailing **Ungrouped**
+section, each sorted alphabetically by name. There's no column sort here,
+since the point of this view is the roster, not ranking. Each row shows
+sessions, tokens, estimated cost and when that person last synced. Click a
+name to open their own activity page: the same Activity/Projects/Tools/
+Health view the single-client Argus app shows (see [Overview](/overview),
+[Metric Views](/metric-views)), scoped to their usage alone.
+
+## Groups
+
+Groups are how you scope Argus Hub's other views (Activity, Tasks, Tools) to a
+team rather than the whole organization: a department, a pod, whatever
+division makes sense for you. Group names must be unique on an Argus Hub.
+
+- **Create a group** from the button above the table, giving it a name.
+- **Rename** or **delete** a group from its section header. Deleting a
+ group doesn't remove its members, it only ungroups them, moving them
+ back to the **Ungrouped** section. Nobody's sync history or usage is
+ affected.
+- **Move one person** with the row action next to them: a popover lists
+ every group, lets you search or create a new one on the spot, and
+ applies it immediately. **Clear** removes them from any group.
+- **Move several at once** by selecting their checkboxes (across sections)
+ and using the bulk action bar that appears, choosing a target group or
+ **Ungrouped**.
+
+Once at least one person has a group, the Group filter appears on
+[Activity](/argus-hub/activity), [Tasks](/argus-hub/tasks) and [Tools](/argus-hub/tools),
+plus an **Ungrouped** option for anyone not yet assigned.
+
+
+
+
+
+
diff --git a/docs/argus-hub/tools.md b/docs/argus-hub/tools.md
new file mode 100644
index 00000000..d53dc2d0
--- /dev/null
+++ b/docs/argus-hub/tools.md
@@ -0,0 +1,90 @@
+---
+description: Which skills, tools, MCP servers and plugins the organization reaches for, aggregated across everyone syncing to an Argus Hub.
+---
+
+# Argus Hub: Tools
+
+Argus Hub's Tools view covers the same ground as the single-client
+[Tools view](/metric-views#tools), skills, tools, MCP servers and plugins,
+aggregated across everyone syncing to the Argus Hub, plus a handful of sections
+that only make sense once you're looking at more than one person's usage.
+
+The same filter bar as [Activity](/argus-hub/activity) applies here: date range,
+source and group.
+
+
+
+
+
+
+
+## Access layer overview
+
+A stat row for the window: how many distinct **tools**, **skills**, **MCP
+servers** and **plugins** were used at all. A note below it calls out what
+share of tool calls the top three tools account for, and how many MCP
+servers make up the long tail beyond the heavily-used ones.
+
+## What people aren't using
+
+A ranked table of the least-reached tools, skills and MCP servers, by call
+count and by how few distinct people ever use them. Unlike the
+single-client Tools view, Argus Hub only sees what's actually invoked. It has no
+visibility into what's installed but never touched, so this section is
+strictly about low or single-person usage, not an enabled-but-unused
+comparison.
+
+## MCP servers
+
+Two charts: calls per [MCP server](/terminology#mcp-server), and the
+heaviest tool results by approximate tokens returned into context. A
+server with high calls and a low tokens-per-call is cheap and popular; one
+with a high tokens-per-call and few distinct users is likely flooding one
+person's context, and worth a look before anyone else picks it up. MCP
+tools render as `server · tool`, so you can tell which server a heavy
+result came from at a glance.
+
+## Tool friction
+
+A table of tools paired with an unusual rate of non-normal stop reasons
+(errors, timeouts and the like). It only appears once enough of the
+window's sessions report a stop reason to make the rate meaningful, since
+a handful of sessions would make any single tool look worse than it is.
+
+## Skills
+
+The organization's top [skills](/terminology#skill) by tokens, plus how
+that use breaks down over time. The table underneath lists each skill's
+invocations, distinct users and sources, which is what separates a skill
+the whole team leans on from one person's habit.
+
+## Tools
+
+The organization's top [tools](/terminology#tool) by call count, grouped
+by category, alongside a full table of every tool with its calls,
+sessions, users and sources.
+
+## Plugins
+
+Argus Hub can only report a plugin as **used** or **not observed** in the
+window. It runs server-side with no local install to inspect, so it can't
+tell you a plugin is enabled but unused the way the single-client view
+does. See [Tools](/metric-views#tools) for that distinction; Argus Hub can only
+show what actually got used, and by how many people.
+
+## Shared vs. solo
+
+Which tools, skills and MCP servers are used by three or more distinct
+people, versus used by only one. Like the per-user rankings elsewhere in
+Argus Hub, this stays hidden until the organization has enough people syncing
+that singling one out wouldn't be identifying. It's a quick way to spot a
+tool worth promoting to the whole team, or one that's meant to stay
+personal.
+
+## Comparing sources
+
+Once more than one [source](/terminology#source) is syncing, a side-by-side
+comparison of tool-category mix and top tools, skills and MCP servers by
+source, since raw tool names differ across agents (Claude Code's `Read` and
+`Bash` versus Codex's `read_file` and `run_shell_command`, for example).
+Categories give you a fair comparison where names don't.
diff --git a/docs/images/screenshots/argus-hub-activity@1920x1080@2.webp b/docs/images/screenshots/argus-hub-activity@1920x1080@2.webp
new file mode 100644
index 00000000..ad0824b8
Binary files /dev/null and b/docs/images/screenshots/argus-hub-activity@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-activity@1920x1280@2.webp b/docs/images/screenshots/argus-hub-activity@1920x1280@2.webp
new file mode 100644
index 00000000..296f8c03
Binary files /dev/null and b/docs/images/screenshots/argus-hub-activity@1920x1280@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-dashboard@1920x1080@2.webp b/docs/images/screenshots/argus-hub-dashboard@1920x1080@2.webp
deleted file mode 100644
index c80ea4ca..00000000
Binary files a/docs/images/screenshots/argus-hub-dashboard@1920x1080@2.webp and /dev/null differ
diff --git a/docs/images/screenshots/argus-hub-export@1920x1080@2.webp b/docs/images/screenshots/argus-hub-export@1920x1080@2.webp
new file mode 100644
index 00000000..c2e8c2be
Binary files /dev/null and b/docs/images/screenshots/argus-hub-export@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-export@1920x1280@2.webp b/docs/images/screenshots/argus-hub-export@1920x1280@2.webp
new file mode 100644
index 00000000..f1b04db8
Binary files /dev/null and b/docs/images/screenshots/argus-hub-export@1920x1280@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-labels@1920x1080@2.webp b/docs/images/screenshots/argus-hub-labels@1920x1080@2.webp
new file mode 100644
index 00000000..3908e628
Binary files /dev/null and b/docs/images/screenshots/argus-hub-labels@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-labels@1920x1280@2.webp b/docs/images/screenshots/argus-hub-labels@1920x1280@2.webp
new file mode 100644
index 00000000..59629e2d
Binary files /dev/null and b/docs/images/screenshots/argus-hub-labels@1920x1280@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-tasks@1920x1080@2.webp b/docs/images/screenshots/argus-hub-tasks@1920x1080@2.webp
new file mode 100644
index 00000000..9e0ba9cb
Binary files /dev/null and b/docs/images/screenshots/argus-hub-tasks@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-tasks@1920x1280@2.webp b/docs/images/screenshots/argus-hub-tasks@1920x1280@2.webp
new file mode 100644
index 00000000..6f8820c7
Binary files /dev/null and b/docs/images/screenshots/argus-hub-tasks@1920x1280@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-team@1920x1080@2.webp b/docs/images/screenshots/argus-hub-team@1920x1080@2.webp
new file mode 100644
index 00000000..ea7f2123
Binary files /dev/null and b/docs/images/screenshots/argus-hub-team@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-team@1920x1280@2.webp b/docs/images/screenshots/argus-hub-team@1920x1280@2.webp
new file mode 100644
index 00000000..0da919be
Binary files /dev/null and b/docs/images/screenshots/argus-hub-team@1920x1280@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-tools@1920x1080@2.webp b/docs/images/screenshots/argus-hub-tools@1920x1080@2.webp
new file mode 100644
index 00000000..5c5ca2e6
Binary files /dev/null and b/docs/images/screenshots/argus-hub-tools@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-tools@1920x1280@2.webp b/docs/images/screenshots/argus-hub-tools@1920x1280@2.webp
new file mode 100644
index 00000000..280b3794
Binary files /dev/null and b/docs/images/screenshots/argus-hub-tools@1920x1280@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-user-activity@1920x1080@2.webp b/docs/images/screenshots/argus-hub-user-activity@1920x1080@2.webp
new file mode 100644
index 00000000..22a3ede0
Binary files /dev/null and b/docs/images/screenshots/argus-hub-user-activity@1920x1080@2.webp differ
diff --git a/docs/images/screenshots/argus-hub-user-activity@1920x1280@2.webp b/docs/images/screenshots/argus-hub-user-activity@1920x1280@2.webp
new file mode 100644
index 00000000..a1539cea
Binary files /dev/null and b/docs/images/screenshots/argus-hub-user-activity@1920x1280@2.webp differ