Workflows, agents, connections and settings
Where you find every workflow and what it's made of, the connections it uses, how much it costs, and your organization's settings.
Updated October 2, 2026 · 6 min read
Workflows
Workflows lists every process your organization designed: its project, stage (Draft, Staging, Production), health, where it runs, the tools it uses, its last run, the share of the last 30 days' runs that succeeded, the average time and this month's tokens.
Filter by project, stage, health, a tool it uses or its name; the filters stay in the address. New workflow opens the same choices as the Dashboard's Start something.
The project chosen at the top of the page (next to the breadcrumb) filters Workflows, Agents, Skills, Runs and Usage together, and this browser remembers it.
A workflow's page
Click a workflow to see what it does and how it's doing, without opening the Canvas. Run opens its Canvas with the run panel ready; Edit in Canvas opens the editor.
| Tab | What it shows |
|---|---|
| Overview | What it does, the connections it needs and whether they work, what Pre-Flight finds, its recent runs and its key facts. |
| Flow | How work moves: as a graph, or as numbered steps that say who receives the task, what each agent uses, where it hands work on and what each approval step waits for. The last run's outcome is shown on each step. |
| Runs | Its runs, with the same filters as the Runs page. |
| Connections | The workflow on the architecture map: its agents, the tools they reach and what crosses to each one. |
| Usage | This month's tokens, runs and estimated cost, by agent, and a link to the Canvas's token heatmap. |
| Docs | The Co-Pilot's design brief when the Co-Pilot designed it. |
| Settings | Its stage (changed in the Canvas), its project, its public app, and Delete. |
Deleting a workflow also unpublishes its public app; its past runs stay in Runs. The page asks first and lists what depends on it.
Agents
Agents lists every Root Orchestrator and Sub-Agent across your workflows, so you can find, for example, every agent that uses Notion. Each row says what the agent is responsible for, its engine, its reasoning and output (Extended Thinking, Artifacts), its work method and the tools it reaches.
An agent's page shows its configuration, its Skills and tools with their connections' state, who hands it work and whom it hands work to, and its part of the last runs. Agents are changed in the Canvas: Edit in Canvas opens the workflow with that agent selected. Save to library keeps a copy to reuse; copies added from the library are grouped on the page.
Skills
Skills lists every action your agents can take: what it does, its inputs, the connection that serves it and whether that connection works, and the agents that use it.
A Skill without a Connector says Needs a connection to run: it can be designed and simulated, but a Live run can't call it until its Connector is added in the Canvas.
Prompts & context
The instructions and knowledge your agents get, in three tabs:
- Library: reusable prompts, notes and saved agents, with their size in tokens and, for a saved agent, the workflows its copies are in.
- Company Memory: every Reference across your workflows, where it comes from (a path, an address, or text written in the workflow), how the AI is told to read it, and which agents read it.
- Co-Pilot memory: the facts the Co-Pilot remembers about your organization between designs (moved here from the Dashboard).
Community & Solutions
Where your organization has sharing switched on, Community & Solutions takes the place of Templates, with four tabs:
- From the platform: the templates, as before.
- From your company: workflows people in your organization shared. Search, filter by category or by what your connections already cover, then View, Use or Save.
- My contributions: what you shared, drafts and withdrawn ones included.
- My collections: your own lists of solutions, visible to you only.
Sharing starts on a workflow's page with Share in your company: describe it, then review exactly what goes with it (what's taken out, which References' text you include, its Context Pack) before you publish. Each version stays as it was published. You can let others publish their own versions; theirs then names you and your solution.
Using a solution makes a new workflow of yours after a review: it uses your connections and capability switches, keeps every approval step and never runs by itself. Make your own version appears when the author allows it. A copy made with Use tells you when a newer version is published; it never changes by itself.
A solution says Tested in Simulation or Tested with a live integration only when someone in your organization ran an unchanged copy successfully. On its page, say It worked or Report a problem: only the author and admins see the counts.
The Community (where your organization has it on) lists solutions other organizations shared and our reviewers approved. Using one makes a workflow of yours after you accept its terms. Your admins submit your company's solutions from their page: they tick that they may share it, choose whether your organization's name shows, and our reviewers approve it before anyone else sees it. Only numbers come back, never who.
Connections
Connections lists what your organization connects to: the Claude API and your tools. Each row shows its status, its route (calls go from the platform's server today), what it may do (the sign-in permissions the platform asks for, or what your token allows), how many workflows use it and its last test. Counts by status at the top filter the list.
Tools a workflow needs but nobody connected are listed next, then the catalog to connect a new one. Organization map shows everything on the architecture map.
A connection's page is its fact sheet: what it reaches, how it signs in, the server runs call, which workflows use it (through a Skill, or all of its tools through a Connector linked straight to an agent), the credential's name and when it was replaced (never its value) and the last test. Test, Reconnect (also how you replace a credential) and Remove, which lists what depends on it first.
Usage
Usage shows a month of Live usage: runs, input and output tokens, Extended Thinking, and the cost estimated at the API's list prices (an estimate, not your invoice). Your plan's limits (Live runs, workflows, connections) are shown as meters.
The daily chart has a table view. Break the month down by project, workflow, agent, provider and route, or engine. Simulation never calls the AI and never counts against your plan; it's listed apart.
Reports
Reports are made when you open them from what the platform records; nothing is stored. The Usage report covers a month for the organization or one project. Print it or save it as a PDF (it always prints light), or download it as CSV or JSON.
Settings
| Section | What it holds |
|---|---|
| Organization | Its name, your role, how many members it has and the Co-Pilot memory switch (admins). |
| Members & roles | Who is in the organization, with their role and status. Admins invite by copying a link bound to one e-mail (it works for 7 days, only for that e-mail's Google or Microsoft account), change roles, suspend and remove; the organization always keeps one active admin. Members and open invitations take the plan's seats. If you belong to several organizations, switch in the account menu. |
| Permissions | What each role can do: the same table the server checks. |
| Plan & billing | Your plan, what it has used, and Stripe. |
| Preferences | Theme, language, table density (comfortable or compact) and voice, for this browser. |
| Data & privacy | What the platform keeps and what this browser keeps, and the Trust Center. |