What is in a workspace
A workspace is the unit of work on this platform, and its console carries the whole platform scoped to that boundary. This page is the catalog of what is in it, grouped the way the console groups it.
Two things are worth knowing before the list.
Nothing here is a placeholder. Every tool below is a live surface backed by a real service. There is no "coming soon" tab in the workspace console.
A clean room is the same console. The clean-room console mounts this entire tree and adds a governance layer on top. See clean-room governance for what it adds and what it changes.
How a tool appears or does not
Four independent layers decide whether you see a tool, and they answer different questions:
| Layer | Question | Effect when it fails |
|---|---|---|
| Permission | Is this principal allowed? | The tool is not reachable |
| Capability | Is this licensed for this workspace? | The tool is not rendered at all |
| Experience | Does this persona need it? | Hidden from this view, still reachable by pin or by switching |
| Backend scope | Is this principal a member? | The call refuses, deny-closed |
Only the third is cosmetic. An experience is a persona filter, not a security boundary, and the platform is explicit about the difference: hiding a tab in a role view is a convenience, and the enforcement is always the layers underneath.
Overview
| Tool | What it does |
|---|---|
| Overview | The workspace at a glance |
| Feed | The workspace's own governed feed, access-filtered server-side and failing closed |
Data and catalog
| Tool | What it does |
|---|---|
| Data | The workspace data plane: engines, databases, tables and lakehouse, scoped to the catalogs bound here |
| Catalog & lineage | Bound catalogs with their lineage, and the path to promote an asset into the marketplace |
| Catalog bindings | Bind a catalog into this workspace, change its privilege between read and read-write, or unbind it. Binding is symmetric and explicit rather than inherited |
| Semantic | Semantic models and metric definitions, the layer a MEASURE() call resolves against, with import from existing metric definitions |
| Data modeling | Transformation projects in dbt's vocabulary: projects, models and seeds, with ref() resolution and real table, view and incremental materializations. Distinct from running an actual dbt project, which goes through a connected repository. See data modeling |
| Ontology | The bound ontology database with live type, link and action counts, and industry-pack installation reporting what was created, skipped and what conflicted. Binding an existing database is supported; provisioning a new one from here is on the roadmap |
| Models | The models granted to this workspace. A workspace consumes models; it does not author them |
| Dimensional | Dimensional design: dimensions with their slowly-changing type and grain, hierarchies and measures, with a live preview. Models persist, created on first save and updated thereafter |
Ingestion
| Tool | What it does |
|---|---|
| Sources | External object-store and warehouse connections, with a health check run at connect time and a one-click path from a source into a query notebook |
| Pipelines | Stream loaders, declarative pipelines, the pipeline composer, the run history, and notebook-backed pipeline steps |
Authoring
| Tool | What it does |
|---|---|
| Notebooks | Polyglot notebooks across Python, SQL, graph and Gremlin. See notebooks |
| SQL editor | Cross-engine SQL with a results grid and query history. The only in-console code-editing surface |
Intelligence
| Tool | What it does |
|---|---|
| Advisor | A conversational assistant bound to this workspace and scoped by the same data-security layer, so it cannot see what you cannot |
| ML | Feature groups, model versions and experiments, with suggested features derived from what the estate scan found |
| Cognition | Scan a source into entities and relationships in the bitemporal knowledge graph. Three honest states: the capability is unavailable, it is off, or it is on |
| AI manifold | Coherence maps over the workspace: which domains glue together, projected per cell |
| Workspace mind | The workspace's fused standing: its operating mode, its net potential, its coherence, and the knowledge it draws on |
| Knowledge | Canonical entities and unresolved contradictions, with an explicit opt-in before this workspace contributes to the node's knowledge |
| Dashboards | Pre-built and pack-materialized dashboards |
| External agents | Register, inspect, refresh and disable external agent peers. Delegation to them runs through a governed, value-gated tool rather than a direct call |
| Agents & skills | The skills bound into this workspace. Read-only by design: a workspace consumes capabilities, and authoring happens in Studio |
| Coding | Submit governed coding tasks and track their runs, choosing the team formation, mode, model and budget |
| Repositories | Browse the git engine: repositories, branches, commit log and per-file changesets. Connecting one is a governed write, and credentials are accepted as a secret reference only, never as a value |
| CI/CD | Continuous-delivery runs, step results, approval gates, promotions and rollbacks. A rollback requires a stated reason |
Operations
| Tool | What it does |
|---|---|
| Compute | Kernels and sessions, with a confirmed stop |
| Capacity | Per-resource gauges, the workspace's share of its group, cruise versus a maintained target, banked headroom, burst standing, and the compensation actions offered under contention |
| Automations | Reactive rules with enable, disable, dry run, and the firing history. See automations |
| Edge sources | Live presence of edge agents streamed rather than polled, connect, disconnect and hold, per-device audit, collectors, data flow and security |
| Logs | The durable error and audit-relevant log tail that survives file rotation, scoped by membership rather than by licence |
| Health & drift | Structure-engine self-awareness: filter false-positive rates, stuck streams, tree depth and inventory, each with a warning line kept in sync with the engine's own defaults |
| Residency | Where data and compute may live, the resolved cells with an inside or outside-boundary flag, and proof that a failover stays inside the boundary |
| Activity | The hash-chained event timeline for this workspace: creations, materializations, queries, shares, clones and governance denials |
| Agent observability | The live agent and sub-agent run tree, signed transcripts and spend rollups, with trends and run comparison. See observing an agent |
| Jobs | Run, list and cancel job runs with a live journal, and create, enable, disable and delete scheduled definitions. See agent jobs |
Consulting
| Tool | What it does |
|---|---|
| Consulting | The client-estate engagement surface: connections, findings with state and trend, cost and savings, data security, exposure, identity, migration planning, storage, virtualization, infrastructure-as-code export, and deliverable approval. See consulting |
Security
Each of these is a real security capability scoped to the workspace, gated on both a permission and a licensed capability.
| Tool | What it does |
|---|---|
| Detections | Detection rulesets bound to this workspace |
| SOAR | Response playbooks bound to this workspace |
| Threat intel | Indicator and campaign feeds |
| Forensics | Forensic cases |
| Posture | Posture and compliance assessments |
| SIEM | Event-lake instances and the events in them, read from the durable lake |
| UEBA | Entity behaviour: an insider-threat score with its risk band, the weighted factors behind it, the baseline sample count, alerts and watch state |
| DLP | Data-loss-prevention policies by category, minimum confidence, action and scope |
| Deception | Honeypots, canaries and decoys |
| Red and purple team | Adversary-simulation campaigns |
| Threat detection | Botnet and command-and-control campaign tracking |
Mission
The most restricted tools, gated on a licensed tier and export control on top of everything else.
| Tool | What it does |
|---|---|
| Cyber operations | Missions with status, phase, stealth posture, cell and scan assignments |
| Battlefield operations | Fire missions with engagement status, firing unit, weapon system, target, rounds and commander authorization |
| Surveillance | Perception feeds resolving observed to candidate to canonical identity |
| Spectrum operations | Radio plans and the mission communications readout |
Governance
| Tool | What it does |
|---|---|
| Data products | The products this workspace ships, and its publish reach: a sharing-tier ladder where each tier above private maps to its own capability, and a locked tier shows the reason it is locked rather than disappearing |
| Data exchange | Discover, offer, manage, earnings and spend. Request, approve and revoke governed access, and publish as a data product, a workspace share, or over an open sharing protocol |
| Governance | Entity voting and certified elections, scoped to this workspace |
| Members & access | Membership and roles, alongside the resolved capability set for this workspace |
| Storage & SFTP | Namespaces, managed tables, buckets, keys, presigned URLs, tiering, retention and integrity, plus the access-control list. SFTP access is governed by path grants |
| Lifecycle | Clone, archive and restore, disaster recovery, a provisioning assistant, and the activity record. See lifecycle |
| Settings | Name, description, storage limit, and deletion behind a confirmation |
The pattern across all of them
Three properties repeat, and they are the reason the console holds together:
Consume here, author elsewhere. Models, skills and agents are bound into a workspace and listed read-only, with a link out to where they are authored. A workspace does not become a second authoring surface with its own governance.
Gate, do not omit. Where a tool is unavailable, the platform prefers showing it with the reason over hiding it, so the navigation tells the truth about what exists. Hiding is reserved for the persona filter, which is not a security boundary and does not pretend to be.
Scope is resolved beneath the call. Every tool resolves the workspace from the console rather than trusting a parameter, which is why the same component can serve a workspace and a clean room without forking.
Related topics
- The boundary itself: workspaces and clean rooms.
- What a clean room adds on top: clean-room governance.
- Which tools each persona sees: experiences.
- Choosing between the two boundaries: choosing a boundary.