Workstreams
Business Summary: Workstreams give operators and leaders one place to pursue a business goal — with the right AI team, connected data, human approvals, and a durable record of what was decided.
Each workstream is a scoped workspace tied to a single goal, specialist team, connected data sources, approvers, and a full history of runs and decisions.
On this page
- How this fits in Nimbus
- Start here
- What is a workstream?
- Workstream workspace and settings
- Checkpoints and approvals
- Creating from chat
- Workstream teams and access
- Workstream Studio and process panel
- Collaboration rail and Runtime
- Recommended path
- Example scenarios
- FAQs
How this fits in Nimbus
flowchart TB
Connectors[Connectors]
Wiki[Company Wiki]
Workstream[Workstream]
Governance[Governance]
Memory[Operating Memory]
Connectors --> Workstream
Wiki --> Workstream
Governance --> Workstream
Workstream --> Memory
Start here
| Goal | Guide |
|---|---|
| Browse active and archived work | Workstreams hub |
| Create via Perception wizard | Creating a workstream |
| Analyze, Plan, Action, Reports + collaboration rail | Workstream chat modes |
What is a workstream? {#what-is-a-workstream}
A workstream is a scoped execution workspace where you pursue one business goal with a governed agent, connected data, a team, and a full audit trail — a dedicated room for one initiative, not a chat thread that disappears when the window closes.
An agent is a reusable specialist team; a workstream is a specific assignment using that team (for example, "Q3 demand forecast for EMEA" on a Finance agent). Each workstream keeps private chat per user, plus shared outputs, milestones, NTU spend, and approvals visible to the team.
| Element | Purpose |
|---|---|
| Goal | The decision or outcome you are driving toward |
| Agent | The specialist team assigned to the work |
| Connectors | Intelligence and action integrations bound to this work |
| Team | People who can access, review, and approve |
| Chat | Private per user — Analyze, Action, Plan, or Reports modes |
| Outputs / Activity | Shared artifacts and audit trail of runs and governance |
Lifecycle: Scope → Commit → Execute → Complete → Archive. Workstreams appear on the hub, sidebar, Overview, and Lifecycle Graph. They inherit governance, meter spend in NTUs, and surface Needs you or Sign-off when human input is required.
Workstream workspace and settings {#workstream-workspace-and-settings}
The workspace has main tabs (workstream chat, Outputs, Activity), the collaboration rail on the right during chat, and header icons (Runtime, Settings). The critical distinction:
| Surface | Purpose |
|---|---|
| Settings (gear) | Configure team, connectors, governance, workflow |
| Collaboration rail | Team-visible plans, evidence, HITL, and committed artifacts |
Settings tabs: Overview (goal metadata), Team (members and approvers), Connectors (bind org integrations), Governance (HITL rules), Workflow (process hooks).
Main tabs: Workstream chat (default Analyze mode) with mode picker and @ attachments. Outputs holds durable artifacts. Activity is the audit trail for runs, NTU spend, and governance events. Runtime (header, left of gear) shows the orchestrator graph during Plan runs.
Checkpoints and approvals {#checkpoints-and-approvals}
Checkpoints pause agent work when a human decision is required — draft review, connector write approval, or high-value spend sign-off. When a checkpoint opens, you see Needs you in chat or Overview, take Sign-off (approve, request changes, or reject), and Activity records the decision.
| Tier | When it applies | Blocks? |
|---|---|---|
| Soft HITL | Clarification, low confidence | Usually no |
| Hard HITL | Draft output before publish | Yes |
| Critical HITL | External writes, high spend | Yes — may need two approvers |
Governance presets (Standard, Strict, Custom) set approver count and gate types at creation or in settings. Act on checkpoints in HITL review on the collaboration rail — not in Settings.
Creating from chat {#creating-from-chat}
Turn a Perception conversation into a committed workstream when your goal is clear. Perception (global button) explores wikis and connectors without auto-running specialists; workstream chat inside a committed workstream supports governed modes.
| Situation | Best path |
|---|---|
| Known name and objective | Scoping form |
| Exploring context first | Perception chat, then commit |
| Thread already captured goals | Ask to create from chat |
Flow: describe goal → Nimbus infers agent and connectors → review preview card → confirm (Goal committed — workstream record created) → open workspace. No record is written until you approve the preview.
Workstream teams and access {#workstream-teams-and-access}
Access combines organization role (Admin, Member, Viewer) with team membership (which workstreams you can open). Both apply. Workstream roles: Owner, Approver, Contributor, Viewer.
Workstream-only access limits visibility to assigned teams — useful for sensitive finance, legal, or M&A goals. Assign teams during creation, in settings, or via admin. Include at least one approver for Hard or Critical checkpoints; Strict governance requires two.
Workstream Studio and process panel {#workstream-studio-and-process-panel}
Workstream Studio (header Open studio) adds a memo beside a live process flowchart — an orchestration command center where you work with the agent orchestrator while sub-agents handle extraction and connector ops.
Right rail tabs: Memo, Notebook, Deliverables, Checkpoints, Process (flowchart with Fullscreen), Settings. Chat with data at the bottom queries attached datasets without leaving the memo.
| Need | Use |
|---|---|
| Quick chat | Workspace workstream chat |
| Memo + flowchart | Workstream Studio |
| All workstreams lifecycle | Process Board (Collaboration sidebar) |
Note: If you only see four workspace tabs, use workstream chat and Activity for the same work — studio adds flowchart layout when available.
Collaboration rail and Runtime {#collaboration-rail-and-runtime}
Inside an open workstream, the collaboration rail (right-hand accordion) tracks Gather evidence, Create plans, Make reports, HITL review, and Commit artifacts — all visible to workstream team members. It is not the gear Settings panel — Settings configures team and connectors; the rail monitors shared execution and approvals. Chat on the left stays private to each user — see Private chat vs shared work.
Open Runtime from the header (left of the gear) to see the orchestrator graph during Plan runs. Full walkthrough: Collaboration rail and Runtime in Workstream chat modes.
Recommended path
| Step | Action | Guide |
|---|---|---|
| 1 | Open the hub | Workstreams hub |
| 2 | Create your first one | Creating a workstream |
| 3 | Pick the right chat mode | Workstream chat modes |
| 4 | Use the collaboration rail | Workstream chat modes — Collaboration rail |
Example scenarios
Q3 baseline with Hard and Critical checkpoints
A finance team runs Plan mode to produce a Q3 baseline memo. The agent triggers Hard HITL; the FP&A lead requests a chart revision, approves, and the memo publishes to Outputs. A Critical checkpoint gates a proposed ERP write; the controller signs off. Activity records both approvals with timestamps for audit.
Perception exploration to committed workstream
An operations lead opens Perception, explores supplier lead times from last quarter, then says, "Create a workstream to recommend safety stock changes for board review." Nimbus proposes EMEA safety stock review, assigns the Supply chain agent, and shows a preview. She confirms; the workstream appears on the hub. She switches to Plan mode with NTU quote and checkpoint sign-off — without retyping scope in a form.
FAQs {#faqs}
Q: Do I need a workstream to run an agent? A: Yes for governed, auditable work. A workstream binds goal, agent, connectors, team, and history in one place.
Q: Where do I create a workstream? A: New workstream on the hub opens the Perception wizard, or commit from Perception chat. See Creating a workstream.
Q: Where do I see items waiting for my approval? A: HITL review on the collaboration rail — see Workstream chat modes.
Q: What is the difference between an agent and a workstream? A: An agent does the work. A workstream is the goal that work belongs to — chat, connectors, outputs, and history for that goal.
Q: Is Settings the same as the collaboration rail? A: No. Settings configures team and connectors. The rail monitors execution and approvals during workstream chat.
Q: Is Perception chat the same as workstream chat? A: No. Perception is workspace-wide exploration. Workstream chat supports Analyze, Action, Plan, and Reports inside a committed workstream.
Q: I cannot open a workstream my colleague sees — why? A: You are likely not on the assigned team or your org tier is guest/viewer. Ask the owner or an admin to add you.
Q: Do Analyze mode charts need approval? A: Usually not. Formal Reports and connector writes trigger Hard or Critical checkpoints.
Q: What is the difference between checkpoint and milestone? A: A milestone marks progress. A checkpoint is a governance pause requiring sign-off.
Q: Is the Process panel the same as Process Board? A: No. Process panel = one workstream flowchart in Studio. Process Board = all workstreams in kanban columns.
Q: Can I reopen a completed workstream? A: Yes. Completed workstreams remain accessible for review and follow-up without losing history.
Q: Can teammates read my workstream chat? A: No. Chat is private to each user. Teammates see plans, runs, reports, and rail artifacts — see Workstream chat modes — Private chat vs shared work.
Q: Do I need connectors before creating a workstream? A: Not always, but analysis and action modes work best once connectors are bound in Settings.