Dashboard, runs and projects
Where you see what's happening, what needs you, how runs ended, and how projects, workflows and providers connect.
Updated October 8, 2026 · 5 min read
The Dashboard
The page you land on after signing in. Every block answers “what's happening” or “what needs me”, and every number opens the list behind it.
| Block | What it shows |
|---|---|
| System status | Whether the platform is normal (admins see which part isn't), your connections' health, whether Claude is connected or you're on Simulation only, and how many runs are in progress. |
| Needs attention | Things a person should look at, most serious first, each with one action. See below. |
| Live now | Runs in progress: the agent working now or the approval it waits at, who started it and since when. Review and Stop are here. |
| Runs · last 7 days | How many runs succeeded, failed or were stopped, the success rate per day, and the latest five. |
| Usage | This month's tokens in Live runs, your plan's Live runs, the workflows that used the most, and an estimate at the API list prices (not a bill). |
| Active projects | Projects that ran something in the last 7 days, with their health. |
| Recent activity | The latest changes: projects, workflows, connections, secrets, approvals, stopped runs. Admins see everyone's; everyone else sees their own. |
Start a Project (top right) opens the Project Starter: describe what you want to accomplish and review a proposed project. Start something next to it adds a workflow: the Co-Pilot, a template or a blank one. Until you have a workflow or a connection, the Dashboard shows a short checklist instead, and its text box starts a project.
Needs attention
Fixed rules, no AI: each item says what happened, where, and the one thing to do. The bell in the top bar shows how many there are.
| Item | When | Action |
|---|---|---|
| Approval waiting | A run waits at an approval step. | Review: read (and edit) the task, then approve or reject. |
| Paused by the loop guard | An agent made too many tool calls in a row. | Review: resume or stop. |
| Run failed | A run failed in the last 24 hours (one item per workflow). | Open the run: what failed, where, and how to fix it. |
| Connection failed or limited | A connection's last test failed or passed with limits. | Test it again on Connections. |
| Connect Claude / Claude rejected your key | There's no Claude key for Live runs, or Claude refused it. | Connections. Simulation works without a key. |
| Plan usage | Most of this month's Live runs are used, or all of them. | Usage, or see plans. |
| The Vault is unavailable | The server can't open saved credentials. | How to fix (for whoever runs the server). |
Runs and the run page
Runs lists every run, Live and Simulation, newest first, with the ones in progress pinned on top. Filter by period, project, workflow, outcome and mode; the filters stay in the address, so a link opens the same list. Export saves the current list as CSV or JSON (up to 500 runs).
A run page shows how it ended, a timeline of the agents and approval steps it went through (outcome, engine, tokens, time, tool calls, waiting time at each approval and whether the task was edited), what failed and how to fix it, and the activity about it (who stopped, resumed or approved).
Projects
A project groups the workflows of one area of work. Workflows you saved before projects existed are in a default project; New workflow on a project files the new one there, and Move to… on the Workflows tab moves one to another project.
- Overview: a small map, the workflows with their health and last run, recent runs, the connections they use and this month's usage.
- Architecture: the project's map (below).
- Workflows and Runs: the project's own lists.
- Settings: name, description, archive. Members and areas come with company spaces (Phase 11).
A workflow's health: Failing when its last run failed; Needs attention after a failure in the last 7 days or when a connection it uses failed or is limited; Healthy otherwise; No runs yet when nothing ran in 30 days. A project shows its Project health instead: whether it's configured, connected and tested, with a fix for each problem. New project starts one from an objective (the Project Starter) or empty.
The Architecture View
How everything connects, in lanes: Platform (projects, workflows, agents, approval steps, Skills, References) and Providers & services (Claude and the tools you connected). Your environment joins when the Bridge exists (Phase 12).
- Select an item (click, or Tab and Enter): the rest dims except what it's connected to, and the side panel shows its facts, its links and Open.
- Health shows the status of every item; Data boundaries says what crosses to a provider; Last run marks the path of each workflow's last run.
- List shows the same thing as grouped lists, for keyboards, screen readers and phones.
- On the Projects page, Architecture shows the whole organization, with workflows closed: select one and choose Show its agents.
What is kept
To build these pages the platform keeps metadata only: each run's workflow and project, outcome, engine, token counts and times; per agent, the same numbers; per approval step, the outcome, the waiting time, whether the task was edited and who decided; daily totals; and the activity log. Raw run facts and the activity log are kept for 12 months, daily totals for 24. Never the tasks, the answers or what a tool received or returned.