Task guides
Choose the task you need to complete. Start with Get started if this is your first visit. Your workspace, permissions and enabled features determine which controls are available.
Use assistants and apps
- Ask your first question.
- Understand assistant results.
- Use an app through its chat, form or block interface.
- Manage assistant memory.
- Manage your notifications.
- Select a workspace.
Build and test
- Follow the workload journey.
- Create a workload.
- Manage assets and versions.
- Create a workflow or start from a template.
- Test and evaluate changes, including comparisons and their limits.
- Prepare a release for review.
- Activate and manage releases.
- Configure connections and secrets.
- Configure AI resources, including providers, models and the workspace assistant.
Monitor and respond
- Investigate runs and human decisions: find a run, inspect its evidence, assess recovery and respond to a pending human decision.
- Check a run’s identity, environment and current state before acting. Retrying can create new external effects; a recorded decision is not proof that execution succeeded.
Review and govern
- Review a release and record your decision.
- Review governance and evidence: policies, organization commitments, evidence, framework posture and supplier assurance.
- Explore and maintain AI Estate: inventory coverage, systems, ownership, dependencies and impact views.
Administer your organization
- Understand access and permissions.
- Manage workspace access: members, effective permissions and role overrides.
- Understand namespaces and workspaces.
- Inspect and change resource limits: workspace concurrency and organization token/cost limits.
- Configure the workspace knowledge assistant.
Specialist operations and integrations
These tasks need additional technical context and delegated access. The linked references are explicitly technical; they supplement the portal tasks above.
| Task | Where to continue and why |
|---|---|
| Register an external system, agent peer or MCP server | External systems, agent peers, MCP and workloads: integration identities, contracts and governance setup require the responsible integration owner. |
| Propose a governed workload change | Change proposals: specialist review, document and binding contracts apply in addition to normal release approval. |
| Configure automated triggers or embedding | Triggers and runtime integration: schedules, external endpoints and application integration need technical configuration. |
| Investigate containment, quarantine or recovery evidence | Dependency quarantine and execution recovery: involve your authorized platform operator; inventory or a diagnostic view alone does not authorize intervention. |
| Export retained audit evidence or manage data requests | Audit reference and data requests: retention, disclosure and platform access need the responsible governance operator. |
| Configure client applications, provisioning or platform namespaces | Identity reference: credentials and cross-organization setup require identity/platform administration. |
| Diagnose platform health or deploy infrastructure | Observability and Kubernetes delivery: service diagnostics and deployment belong to platform operations. |
Get help
Use Troubleshooting for symptoms and next actions, or the Glossary for unfamiliar terms. Technical documentation lists the complete API, SDK, CLI, advanced configuration and architecture reference. Existing reference URLs remain available.