Features — the catalogue
One row per thing the product does, grouped by where the reader meets it. Who is the smallest role that can use it (roles.md); Where is the screen, named as the Console names it.
If a feature ships and does not appear here, it is not done
(docs/CONVENTIONS.md §8).
Working with a session
Section titled “Working with a session”| Feature | Who | Where | Details |
|---|---|---|---|
| Start a session — agent, model, effort, start mode | member | workspace action bar → Start | Start a session |
| Run in a fresh git worktree | member | Start dialog | Work in parallel with worktrees |
| Chat mirror — follow and steer a running agent | member | main area | Reply to a session |
| Answer a question, plan, or permission prompt | member | mirror | Reply to a session; agents.md for which agents ask |
| Answer a question or plan left by a stopped session | member | mirror, the carried card | Reply to a session |
| Hold follow-ups while a turn runs (Waiting to send) | member | mirror composer → Send ▾ → Add to waiting list, or Alt+Enter | Hold what you send, stop a turn |
| Have another session review a plan | member | the plan card | Comment on a plan, have it reviewed |
| Skill / command picker | member | mirror composer | Call a skill or a command; agents.md for which agents |
| Live terminal attached to a session | member | main area | Use the terminal |
| Terminal round-trip time (keystroke lag) | member | terminal pane header | Use the terminal |
| Ctrl+C / Ctrl+V copy and paste in the terminal | member | terminal | Use the terminal |
| Resume a stopped session | member | left pane → the session | Stop and resume a session |
| Home — what waits for you, where you left off, at a glance, and the first-run setup | member | the empty main area, action bar → Home, or leader g h |
Find your way around the screen |
| Sessions overview — every running session as a card | member | action bar → Sessions, or leader g s |
See every running session |
| Fleet graph — one lane per session, time left to right | member | sessions overview → Graph, or leader g f |
See how sessions relate over time |
| Stop every session under a working copy | member | repository / worktree right-click → Stop the sessions below | Stop and resume a session |
| Hand a conversation to a new session | member | session ⋯ menu | Hand off to another session or agent |
| Move a session to another agent | member | session ⋯ menu → Move to another agent…, or the usage-limit band in the chat | Hand off to another session or agent |
| Export a conversation as Markdown | member | session ⋯ menu → Export as Markdown | Save a conversation as Markdown |
| Fork from a past message | member | mirror, on a past message | Branch from a past message; agents.md for which agents |
| Hand a session to another member | member | session ⋯ menu or handoff proposal card → Hand off to a member | Hand a session to a member |
| Share a session read-only | member | Shared sessions | Share a conversation |
| Highlights | member | mirror, on a selection | Reply to a session |
| Changed files | member | strip under the mirror header | See what a session changed |
| Context usage gauge | member | session header | Check prices and usage |
| Read an answer in your own language (translation, no turn spent) | member | mirror, under an answer | Read an answer in your own language |
Have a session generate an image (generate_image) |
member | any session not attached to a studio, by asking | Have a session make images |
| Start a session from a session (child sessions) — hand a task to another agent, look after what it started | member | Settings → Agents → Session, then any session, by asking | Start child sessions from a session |
| Abort detection and auto-resume | member | automatic | 04 Conversation |
Working with code
Section titled “Working with code”| Feature | Who | Where | Details |
|---|---|---|---|
| Import a repository | member | Repositories → Clone | Clone a repository |
| Start from an empty folder | member | Repositories → new folder | Clone a repository |
| Commit graph, diff, stage and commit | member | Commit graph | Commit in the commit graph |
| Worktrees | member | under each repository | Launch a session from a repository |
Use gh and bkt in a session, already signed in |
member | any session | Use the GitHub CLI, Use the Bitbucket CLI; the list of signed-in CLIs is in Commands |
Use jira in a session, already signed in |
member | any session | Use Jira |
Use confluence in a session, already signed in (the Jira connection) |
member | any session | Use Confluence |
Edit and delete Jira comments (jira issue comment edit / delete) |
member | any session | Use Jira |
| Subversion working copies — update, show log, local changes | member | Repositories → an SVN row | Use a Subversion repository |
| File tree and viewer | member | Files | See and download files |
| Download a folder as zip | member | Files tree or gallery → right-click → Download as zip | See and download files |
| Open a shell from a folder | member | Files tree → right-click a folder → Open shell here | See and download files |
| Links in rendered Markdown — tickets, file paths, loopback URLs that open the workspace preview | member | mirror, chat, file and document viewers | View documents and diagrams |
| Menu on file and folder links in replies and documents | member | right-click a link | View documents and diagrams |
| Markdown and code editing | member | Files → a file | Edit a file |
.drawio diagrams |
member | Files → a .drawio file |
View documents and diagrams |
| PDF and Office document preview | member | Files → that document | View documents and diagrams |
| Image generation — make pictures in bulk without an agent | member | action bar → Images, or leader g i |
Generate images |
| Edit and inpaint — reference pictures, a mask file, instruction-edit models | member | image generation → Advanced → Operation | Generate images |
| LoRAs with their trigger words, shown before you pick one | member | image generation → LoRAs | Generate images |
| Image studio — refine the draft by talking to an agent, a person presses generate (edit history, versions, picture history) | member | image generation → Attach an agent | Finish an image in the studio |
Image knowledge notes — four sections per model and per family, in ~/imagegen-knowledge/ |
member | image generation → Notes, or the file tree | Finish an image in the studio |
| Image gallery — a folder’s pictures as cards | member | file tree → “Open in gallery”, or a session’s “Generated images (N)” | Use the image gallery |
| Gallery card menu — copy, rename, delete, jump to the session that made it | member | gallery card → right-click | Use the image gallery |
| Browser pane for a local web app | member | workspace action bar → Preview | See a web app you started; browser-pane.md for the contract |
| Preview subdomains (minted on every start) | member | workspace action bar → Preview | Use preview subdomains |
| Sign in with Google on a preview URL (development OIDC IdP) | member | Settings → Preview subdomains | Try Google sign-in on a preview |
| Attach to a Chromium the agent owns | member | a link the agent hands you | Operate a browser the agent opened |
Organising the work
Section titled “Organising the work”| Feature | Who | Where | Details |
|---|---|---|---|
| Working sets | member | top of the left pane | Narrow the view with working sets |
| Memo queue | member | Memo queue | 10 Chat and memos |
| Work-item inbox — issues, tickets, pull requests | member | Issue tracker | Start from an issue or a ticket; repos.md for what each provider contributes |
| Scheduled (unattended) runs | member | Schedules | Run at a set time |
| Notification centre | member | top bar | 20 Settings |
| Inbox — messages and requests between members, with claim, launch and PR links | member | top bar → inbox | Ask a teammate through the inbox |
| Inbox recipes — send a prompt with a suggested agent and repository, take it into your memos | member | inbox → New / a card | Ask a teammate through the inbox |
| Attach a session to an inbox message, sharing it with the recipients | member | inbox → New | Ask a teammate through the inbox |
| Red dot on a session with an unread notification | member | left pane, tabs | Read a session’s state |
| Work-item links on a session row — its branch’s pull request and the tickets it is linked to | member | left pane, session row | Start from an issue or a ticket; repos.md for which providers |
| Pull request notifications — ready to merge, merged, closed | member | Settings → Notifications | 20 Settings |
| A session watching a pull request — told when checks finish, conflicts, merges | member | (agent tool watch) |
Tell a session about its pull request’s review |
| Not responding badge — the workspace is up but its Agent does not answer | member | beside the workspace power button | Fix a workspace that will not start or has stopped |
| Engine pills — is the chat / image engine up | member | top bar | Use lcpp |
| Assistant chat | member | Assistants | 10 Chat and memos |
| Chat bridge — Discord / Slack | member | Settings → Chat | Connect Discord and Slack |
| Reply suggestions | member | mirror composer | Reply to a session |
| Deletion lock | member | session ⋯ menu | Clear away finished sessions |
| Cleanup and the trash | member | Clean up | Clear away finished sessions |
| Keyboard system | member | Settings → Keyboard | Use the keyboard and the command palette |
| Text-to-speech | member | Settings → Read aloud | 20 Settings |
| Skills views — each agent’s skills, read-only: personal (Settings) and a repository’s (project tree → Skills…) | member | Settings → Skills; repository row menu → Skills… | Bring your Claude Code setup |
Personal settings
Section titled “Personal settings”Every row is under Settings; the tab is the “Where”. settings.md lists the tabs, 20 Settings explains them.
| Feature | Who | Where |
|---|---|---|
| Agent connections | member | Agents |
| Instructions to agents | member | Agent instructions |
| Agent Fleet shared memory management | member | Agent memory (Read and correct agent memory) |
| Version history of agents’ own memory | member | Memory snapshots (Restore memory from a snapshot) |
| Agent Fleet memory from an agent outside Agent Fleet (off by default) | member | Agent memory → Settings (Use memory from an external agent) |
| Muse Code contributor mode (off by default) | member | Agents → Muse Code → Behaviour (09) |
| Git hosting connections | member | Git hosting |
| Internal repositories | member | Internal repos |
| AWS profiles and SSM | member | AWS profiles/SSM (Use AWS SSO, Log in to EC2 with SSM) |
| Google Cloud profiles | member | Google Cloud (Use Google Cloud) |
| Integration servers and tokens | member | MCP servers (used by sessions) / Outside access → MCP tokens & apps |
| Connect a remote MCP client by signing in (OAuth, when the deployment enables it) — Connected apps, Recent calls | member | Outside access → MCP tokens & apps (Connect a remote MCP client by signing in) |
The “via MCP” badge on input sent through /mcp |
member | input bubble in the mirror (Drive your workspace from an external Claude) |
| Sign in to a registered MCP server that needs OAuth (Sign in, sign in again, approve extra permission) | member | MCP servers (used by sessions) → Authentication → OAuth (Add an MCP server to your sessions) |
| Issue-tracker connections | member | Issue tracker |
| Usage | member | Agent usage |
| Limit board: publish where you stand on usage limits, and see what teammates publish to you | member | Limit board |
| Cloud cost | member | Cloud cost |
| AI assistance (per-feature on/off, agent, model) | member | AI assistance |
| Display, language and keys | member | Display / Keyboard |
| Export / import settings | member | Export / import |
| Toolchains | member | Toolchains |
| Databases (Postgres / MySQL — install, start, stop, reset) | member | Toolchains |
| Preview-subdomain settings | member | Preview subdomains (only where they are issued) |
| Destructive actions | member | Danger zone (Recreate, Clean home; only where the deployment can perform them — deploy-targets.md) |
Tenant administration
Section titled “Tenant administration”Every row is under Tenant settings. admin/ is the shelf; roles.md says what is read-only for whom.
| Feature | Who | Where | Details |
|---|---|---|---|
| Members | tenant admin | Members | Check on your members |
| Deleting a member’s home backups (where the deployment keeps them) | tenant admin | Members → the member | Remove a member |
| Sessions across the tenant | tenant admin | Sessions | See how resources are used |
| Limits and idle auto-stop | tenant admin (read) | Limits & idle | Set session limits, Stop workspaces left idle |
| Limit board on / off for the tenant | tenant admin | Limit board | Set session limits |
| Whether this tenant may use the llm / image engines, and whether its administrators may take models in | deployment admin (super_admin) | Tenant settings → Limits | Choose who may use the inference engines and which models |
| Which of the tenant’s members may use the llm / image engines | tenant admin | Engine access | Choose who may use the inference engines and which models |
| Taking engine models in, when the tenant was granted it | tenant admin | Inference engine models | Choose who may use the inference engines and which models |
| Workspace sizing | deployment admin | — | deploy-targets.md |
| Sign-in methods and login rules | tenant admin | Sign-in methods / Login rules | Choose how people sign in |
| Connection-source restriction | tenant admin | Allowed networks | Limit the networks people sign in from |
| Whether outside apps may connect to the tenant through MCP OAuth | tenant admin | Allowed networks → MCP OAuth | Distribute an MCP server to everyone |
| Integration app OAuth | tenant admin | Integration OAuth apps | Choose the OAuth apps for integrations |
| Branch naming rules — the team’s default branch names and base | tenant admin | Branch naming rules | Set the branch naming rules |
| Member defaults — the agent new members connect | tenant admin | Member defaults | Set member defaults |
| Whether members may use Agent Fleet memory from outside | tenant admin | External memory access | Allow memory use from outside |
| Distributing integration servers | tenant admin | MCP distribution | Distribute an MCP server to everyone |
| Audit | tenant admin | Audit | Read the audit log |
| Running time | tenant admin | Running time | Read running time and cloud cost |
| Cloud cost, per member (AWS only) | tenant admin | Cloud cost | Read running time and cloud cost |
Operating a deployment
Section titled “Operating a deployment”| Feature | Who | Where | Details |
|---|---|---|---|
| Deployment targets | deployment admin | before start | deploy-targets.md |
| Install, upgrade, back up, restore | deployment admin | a shell | Install Agent Fleet / Upgrade / Back up and restore |
| Ingress, TLS and sign-in providers | deployment admin | a shell | Sign-in |
| Audit log and egress control | deployment admin | Admin | Control outbound traffic |
| Which client hosts MCP OAuth trusts | deployment admin (super_admin) | Admin → MCP sign-in hosts | Operate MCP sign-in |
| The development OIDC IdP for preview URLs | deployment admin (super_admin) | Admin → Dev sign-in | Use development sign-in |
| Keys at rest on AWS KMS, and a key of its own for each home | deployment admin | a stack parameter, then a shell | Put the keys at rest on AWS KMS, Give each home its own key |
| The deployment’s colour and label (Appearance) | deployment admin (super_admin) | Admin → Appearance | Tell deployments apart |
| Pausing an idle deployment | deployment admin | a shell | Use idle stop and force-stop |
| Monitoring integrations | deployment admin | Settings → Ops & monitoring | Connect monitoring tools |
| Slot pool and instance classes | deployment admin | Admin | deploy-targets.md |
| Replacing slots after a launch template change | deployment admin (super_admin) | Admin → Slots | Replace and stop slots |
| Force-stopping a wedged slot | deployment admin (super_admin) | Admin → Slots → Stop (force) | Replace and stop slots |
| The inference engines’ GPU class | deployment admin (super_admin) | Admin → Inference engines | Run inference engines on your own GPUs |
| Taking a model in — search, plan card, one press | deployment admin (super_admin), or a tenant admin who was allowed | Admin → Inference engines → Model catalogue | Take models into the inference engines |
| The S3 bucket as the ledger of what is actually held | deployment admin (super_admin) | Admin → Inference engines → Model catalogue → Bucket | Take models into the inference engines |
| Model names, example images and family categories in the catalogue | deployment admin (super_admin), or a tenant admin who was allowed | Admin → Inference engines → Model catalogue → Registered | Take models into the inference engines |
| A repository’s quantisations, with what fits the GPU class | deployment admin (super_admin), or a tenant admin who was allowed | Admin → Inference engines → Model catalogue → Registered | Take models into the inference engines |
| Hugging Face and Civitai tokens for gated downloads | deployment admin (super_admin) | Admin → API tokens | Take models into the inference engines |
| Civitai Red (the NSFW sister domain) as a search source | deployment admin (super_admin), where the deployment was given it | Admin → API tokens | Take models into the inference engines |
| The speech engine and the shared pronunciation dictionary | deployment admin (super_admin) | Admin → Read aloud | Run the read-aloud engine |
| Image generation on a ComfyUI of your own | deployment admin (super_admin) | a shell, then Admin → Inference engines | operator 08 |
| Borrowing another deployment’s llm / image engines | deployment admin (super_admin) | a shell, plus a membership on the far deployment | operator 09 |
| Chat inference on a llama.cpp of your own | deployment admin (super_admin) | a shell, then Admin → Inference engines | operator 10 |
| The user guide inside containers | — | automatic | roles.md |