Getting started in the dashboard
The dashboard helps a new organization get started with a checklist, a demo analysis, and empty pages that explain what to do next. What you see depends on your permissions and deployment:
- the Getting started checklist on the Overview page;
- a demo analysis you can load to explore before running a real analysis;
- empty pages that explain what fills them and what to do next.
The Getting started checklist
Section titled “The Getting started checklist”The checklist at the top of Overview guides you toward your first reviewed product: connect your code, group repositories into a product, choose what to review it on, scan it, and decide on a finding. If you analyze without a code host, the checklist offers a different path. Its title follows your progress, from Set up your first product to Get Checkout reviewed. Each item links to the page where you complete it and explains why it matters. Once a product exists, items name it and link to its tabs. The dashboard checks item status when the page loads; it does not store a “done” flag. For example, deleting your only LLM connection makes that item incomplete again.
An item that depends on an earlier one has no button until the earlier item is done. The card disappears when every required item is done. Hide hides it for the current organization in this browser; Show setup checklist under the page title brings it back. The getting_started_hidden cookie stores this preference (see Cookies and consent).
| Item | Shown to | Only when | Ticked when |
|---|---|---|---|
| Connect your code | holders of integrations:manage | LLM features are on, and the organization has not analyzed without a code host | the GitHub App is installed on a repository, or GitLab is connected |
| Create your first product | holders of products:manage | the organization has repositories, or has not analyzed without them yet | a product holds a repository |
| Choose what it is reviewed on | holders of products:manage | — | a pack is enabled for the organization, a product, or a repository |
| Add an LLM connection | holders of llm_connections:manage | LLM features are on | the organization has a connection |
| See where it stands today | holders of repositories:scan | LLM features are on, and the product path applies | the product has a changeset from a scan or a pull or merge request |
| Run your first analysis | holders of analysis:run | the scan item does not apply to this member | the organization has a job |
| Decide on a first finding | holders of findings:triage | an analysis has not come back empty | a finding is acknowledged, suppressed, or marked false positive |
| Connect a code host (optional) | holders of integrations:manage | the organization has analyzed without a code host | a code host is connected |
| Draw rules from your ADRs and agent files (optional) | holders of repositories:manage | a first product exists | one of its repositories has knowledge files set |
| Name the team that owns it (optional) | holders of products:manage | a first product exists | the product has a team |
| Set a goal (optional) | holders of goals:manage | a first product exists | the product has a goal |
| Invite your team (optional) | holders of invites:manage | — | there is a second member or a pending invitation |
| Notifications for blocked pull requests (optional) | holders of notifications:manage | — | a notification channel exists |
| Configure single sign-on (optional) | holders of sso:manage | — | an SSO connection exists |
| Enable Google or Microsoft sign-in (optional) | platform superadmins | self-hosted | social sign-in is enabled |
| Curate allowed models (optional) | platform superadmins | self-hosted | the models catalog has an entry |
On this dashboard, inviting your team is optional. It is required when the platform runs without the analysis product.
The first product is the oldest one that holds a repository. Items depend on permissions, not role names, so a custom role on the Roles page sees the items its permissions allow. A viewer who cannot complete any checklist items sees no card. Superadmin items do not appear on the operator’s cloud deployment.
Without a code host
Section titled “Without a code host”An organization that analyzes through CI or with LLM features off may have no repositories to group. Once it has run an analysis without a code host, the product items leave the list and Connect a code host moves to the optional items. Connecting a host brings the product items back. Connect your code can also offer Run an analysis for this path.
A member who cannot scan a repository sees Run your first analysis instead of the scan item. If an analysis finds nothing, Decide on a first finding leaves the list because there is nothing to decide.
Trial progress
Section titled “Trial progress”For a cloud organization on a free trial, a line under the title shows trial usage. It includes a See plans link when the operator has set NEXT_PUBLIC_MARKETING_URL.
The demo analysis
Section titled “The demo analysis”Before your organization has analyzed anything, Findings and Changesets offer Explore a demo analysis. The checklist also offers the demo when its first-review step is waiting. Loading it creates a changeset named demo and opens its report.
The demo uses a sample pull request and can include two kinds of findings:
- Agent manipulation. The pull request hides an instruction to the AI reviewer in an HTML
comment, and edits
AGENTS.mdto tell future agents to ignore logging changes. These findings come from the injection scanner, which runs when the demo loads. - Compliance. A pack’s sample can include an invoice export that leaks across tenants, swallows its audit record, and commits a database password. Findings explain which audit framework they relate to. The pack analyzes these findings in advance. When only the Terraform plan pack is enabled, its sample is a plan that replaces the customer-managed key on an invoices bucket, turns off CloudTrail and access logging, suspends versioning, and cuts log retention to seven days.
You can browse and triage findings, open the report, and share it as you would a real analysis. Demo data is inert:
- no LLM is called, so it works before you have a key and with LLM features off;
- nothing is billed, nothing counts against a cloud trial, and no job is created;
- no notification is sent, and triaging a demo finding records no reviewer feedback, so rule health never learns from it.
Demo findings carry a Demo badge. A banner above Findings, Changesets, and the demo’s report explains where they come from. Remove demo data deletes the demo changeset and its findings; the demo is not removed automatically. If agent-manipulation detection is off on the deployment (analysis.injection_scan_enabled: false), the demo contains only the pack’s findings, and the banner says so.
Loading and removing the demo requires the demo:manage permission, held by members and above. See
/demo for the API, and
Adding a pack to ship a sample with your own
pack.
Empty pages
Section titled “Empty pages”Empty pages explain what fills them, how to fill them, and what you can do next. If your permissions do not allow the next step, the page names who can take it and shows no button.
Some pages distinguish “nothing yet” from “nothing matches”:
- Findings and Changesets before the first analysis show No findings yet or No changesets yet, with Run an analysis and the demo when available. If filters match nothing, Findings offers Clear filters instead.
- Analyze shows at most one alert above the form for the first thing that would stop an analysis: no pack enabled for the organization, the cloud trial used up, or no LLM connection.
The navigation lists only pages your permissions allow, so a viewer does not see pages that would return a permission error.
Help bubbles
Section titled “Help bubbles”A (?) next to a page title or field explains the term in a few sentences. Click it to open the bubble, and click elsewhere to close it. Its Read more in the docs link opens the related page on this site.