Runners and Connections
The Run group of the sidebar has Runners and Connections. They answer "where does work run, and what is this organization connected to". Connections also puts a badge on its sidebar item when something needs a look.
Runners
A runner group is where agents and CI jobs run. See Runners for backends. The list splits into Local machines (DevKit groups: a person's own computer runs the work) and Cloud, filtered by All, Working, Idle, Attention, and Search runner groups. The subtitle reads N groups · N live. New Runner Group (⌘N here) opens the create sheet; with no groups the page explains what a group is.
Right-click a group: Copy Group Name, Copy Group ID.
A group has five tabs:
| Tab | Shows |
|---|---|
| Overview | Capacity (how many instances are wanted, the warm pool, an input to change Wanted within the group's minimum and maximum), Activity, 7 days with runs per day and the latest runs, Maintenance and Details (backend, purpose, integration, scope, region, instance type, GitHub runner group, created, UUID) |
| Instances | Instances (N) with state, and Show ended instances for finished ones. Per instance: Drain, terminate, Copy Name, Copy Instance ID, Copy IP Address |
| Runs | Runs (N), filterable, with Stop for a running one and Restart for a failed one (starts a new run). A run opens with its own page and a back button |
| Metrics | Runs per day, Why runs failed, and the Warm pool now |
| Settings | Description, Minimum, Maximum, Runner labels (replaces labels; blank keeps them), Config (JSON merged into the backend config), and Delete |
Maintenance:
| Action | Does |
|---|---|
| Health check | Syncs the group's status and instances from the cloud. Also a toolbar button |
| Drift check | Compares the cloud environment with what the group expects. Read-only |
| Rotate | Drains every instance and lets fresh ones replace them, which picks up a new image |
| Drain | Stops taking new work; busy instances finish first |
Stopping a run or terminating an instance asks first ("The work it is doing is cut off"). On the Overview board, Failing counts groups in error with nothing live and Stale counts instances with no heartbeat for over 6 hours.
To run agents on your own computer, set up This computer under Agents and Coding.
Connections
What the organization is connected to, whether it works, and how to fix it. The list is grouped by provider; Need a look filters to the ones that do, and the sidebar badge counts them. The subtitle reads N connected and N need a look. Search by name.
A connection's page (Check it works now, also in the right-click menu as Check Now) shows its health in words and What you can do:
| Action | Does |
|---|---|
| Rename | A name people recognise when they pick it in a workflow |
| Replace credentials | Paste a new key or token. It is tested before it replaces the old one |
| Sign in again | Renew access by signing in to the provider in your browser |
| Manage the installed app | Change which repositories or resources an installed app can reach |
| Read its | Pick up repositories or resources added or removed since it was connected, for providers that sync |
| Disconnect | Deletes its stored credentials; whatever uses it stops working. A connection that goes away is also told to the assistant, which works from what is connected |
Details lists provider, account, Signs in with (an installed app, signing in with the provider, an access token, an API key, a role it may assume or a service account key), status, Last checked, when it was connected and the connection ID.
Connect something new (the +) lists every provider that can be connected, with Search providers (GitHub, AWS, Slack…). Setup opens a guided sheet per provider; for the platform side of each provider read AWS connections, Cloud connections and DNS providers. Providers that sign in through the browser show Stop while the app waits for you to finish.
Agents and Coding
The Agents section (actors, configs, sessions, costs, teams, this computer) and the Coding section (code hosts, repositories, the coding lane, settings) of Orkestia Cockpit, and how plans across several repositories work
Organization — people, access and cost
Manage people and invitations, API tokens and the cost of an organization from Orkestia Cockpit — roles, one-time secrets, seats, usage, and where payment happens