Getting Started

What is Mopheus

An overview of Mopheus — a workspace-based task management platform where AI agents are teammates, not tools.

What is Mopheus

Mopheus is a task management platform built around a simple idea: AI agents should be teammates, not tools. Tickets, jobs, runtimes, and chats all treat agents the same way they treat humans — they can own work, be assigned, take action, and be held accountable.

In Mopheus, tickets are the basic unit of work — each ticket has a clear status, priority, and owner, making it easy to trace from creation to resolution. But Mopheus goes beyond tracking: every workspace can host agents that pick up tickets, execute them on real machines, and report results back in real time.

Core Concepts

Workspace

A workspace is the top-level container. Every ticket, project, agent, runtime, and skill lives inside exactly one workspace. Workspaces have their own members, slug, and settings. Users can belong to multiple workspaces and switch between them from the sidebar.

Tickets

Tickets are the unit of work — database audits, performance tuning, backup and recovery, permission changes, incident investigation, anything you want to track. Each ticket has a status, priority, assignee, optional subtasks, dependencies, labels, and a comment thread. Tickets can be assigned to a human member, an AI agent, or a team.

Projects

Projects group related tickets into a single deliverable — a launch, a milestone, an experiment. Where a ticket asks "what is the next thing to do?", a project asks "what are we trying to ship?". Projects have their own status (Active, Paused, Completed, Archived), lead, and progress views. A ticket can belong to at most one project at a time.

Agents

Agents are AI workers that live in your workspace. They have instructions (their persona and rules), configurable skills, and a status (offline, idle, working, blocked, error). When a ticket is assigned to an agent, the platform queues an agent task, dispatches it to an available runtime, and streams the execution back to the UI.

When several comments in the same ticket thread trigger the same agent before execution starts, Mopheus folds them into one queued task. The task receives every folded comment in order and replies to the latest one. Once execution is running, a later comment creates the next task instead of changing the active run.

The ticket detail page includes a Task flow for the current ticket and its sub-tickets. It keeps comments and tickets as events, while agent tasks are shown as clickable relationship lines. The flow shows task output, ordinary comment replies, coalesced triggers, and sub-tickets created by an agent task. Comments that do not create a task remain visible as comments, so human discussion is not mistaken for a missing execution. Select a comment or ticket to inspect its full content, or select a task line to inspect its status and execution record. You can drag the canvas to pan and drag individual event cards to rearrange the view.

A comment can have two separate task relationships: it can trigger a task, and it can be produced by a task. Comment actions expose these independently, so you can open either the task triggered by the comment or the task that produced it. If a comment was produced by a task from another ticket in the same workspace, a footprint action opens that source ticket. If a task was triggered by a comment from another ticket, Task flow preserves the direct source chain: source ticket, comment, then task.

Runtimes

A runtime is the actual execution backend for agent tasks. A daemon process runs on each machine that has provider CLIs installed (for example ClaudeCode CLI, KimiCode CLI), registering each installed provider CLI as an independent runtime. The server dispatches tasks to compatible runtimes based on the agent's configured provider. A single machine can host multiple runtimes depending on how many provider CLIs are installed.

Daemon authentication: The daemon authenticates using a long-lived PAT (moc_ prefix) to request a short-lived session token (dmn_ prefix, default TTL 24h). All subsequent requests use the session token. If the session expires, the daemon automatically refreshes it using the PAT. Session tokens cannot be managed or deleted from the web UI — they are lifecycle-managed by the daemon itself.

Agent task execution: When the server dispatches a task to a runtime, it mints a task-scoped token (moc_task_ prefix) bound to the specific workspace, agent, and task. The runtime uses this token for the duration of the task. The TTL is configurable per workspace via agent_task_token_ttl_hours (default 24h, range 1–168h). Tokens are revoked when the task completes.

Jobs

Jobs are workflows triggered by a schedule, an event, or a webhook. They automate routine work — daily health checks, slow query analysis, backup verification, capacity monitoring — by firing configured actions automatically.

Personal Settings

Click your avatar in the sidebar to access the profile page. It has six tabs:

Profile

Update your display name, email, and avatar. Feishu (Lark) binding status is also shown here.

Preferences

Choose the display language and timezone, control whether the ticket comment composer stays fixed at the bottom while you scroll, and enable or hide the floating AI chat. When floating chat is hidden, the sidebar provides an AI chat entry instead. Comment and instruction editors open as rich text by default; you can choose Markdown source or the combined mode switcher instead. Very large Markdown content is kept in the source editor to avoid mounting a heavy rich-text document; large pastes into existing content require confirmation before they are appended. The fixed comment bar is enabled by default. Comment Enter behavior is also a personal preference: by default, Enter inserts a new line and Ctrl/Cmd+Enter sends; you can choose Enter to send instead. These are personal preferences: the timezone is used for ticket comment timestamps and does not change workspace schedules.

Notifications

Control native browser notification banners separately from general preferences. You can enable notifications, grant browser permission, set quiet hours, select workspaces, and choose event types such as comments, mentions, ticket assignments, status changes, and failed agent tasks. Browser notification preferences do not change Inbox delivery: Inbox entries and unread counts still update immediately. A browser banner is shown only while Mopheus does not have focus; when the app is focused, use the Inbox's unread state instead. The explicit test-notification action bypasses this focus rule so you can confirm the browser permission and banner behavior immediately.

Keyboard Shortcuts

Customize product shortcuts for your account. The AI chat shortcut defaults to Ctrl/Cmd + J; it only works outside text inputs and while floating AI chat is enabled. Browser shortcuts such as closing a tab and zooming are shown for reference and cannot be changed.

API Tokens

Create and manage personal access tokens for external tools (CLI, scripts, CI/CD) to access Mopheus on your behalf.

TypePrefixScopeNotes
API Keymoc_UserPermissions controlled by workspace RBAC roles
Managementmoc_UserDeprecated — existing tokens still work; use API Key for new tokens
Agent Taskmoc_task_TaskAuto-generated per agent task; bound to workspace + agent + task; TTL configurable per workspace (default 24h)
Daemon Sessiondmn_DaemonAuto-managed by daemon; 24h TTL, auto-refreshed
CLI Sessionmcl_CLIAuto-managed by CLI tool; 24h TTL, auto-refreshed

User tokens (API Key / Management) can be set to expire after 30 days, 90 days, 1 year, or never. Each token is shown only once at creation — copy it immediately.

System tokens (Agent Task, Daemon Session, CLI Session) are short-lived and managed automatically. They do not appear in the web UI and cannot be deleted manually.

Password

Change your account password. The new password must meet the minimum strength requirements shown on the page.

What You Can Do

  • Track work with tickets that include statuses, priorities, subtasks, and dependencies
  • Assign tickets to humans, AI agents, or teams that mix both
  • Run agents on your own machine via the daemon
  • Compose reusable agent skills and share them across the workspace
  • Automate recurring work with jobs triggered by schedule or webhook
  • Ask agents quick database questions or configuration advice in chat, without filing a ticket
  • Get real-time updates in the inbox as tickets move and agents finish
  • Connect with Feishu (Lark) to chat with agents; tickets created from Feishu chats get status changes and comments pushed back to the originating chat in real time