Accounts and sign-in
Constellation uses one account across its surfaces: the same identity signs you in to the console and to this documentation site. Accounts are backed by Amazon Cognito; you can start as a guest and create an account whenever you want your work and entitlements to persist.
Guest mode
Opening the console without signing in starts a guest session. Guests get the full globe, the demo fleet, and the assistant, with limits:
| Field | Guest value |
|---|---|
| Plan | Guest |
| Chat model | Constellation Basic |
| ML backends | None (upgrade required) |
| Assistant queries | 10 per guest session |
| Fleet | Demo simulation (no live telemetry) |
The account panel (avatar, top right of the console) shows your remaining guest queries, the environment you are on, and the console URL. Reset guest session clears the local session and restores the 10-query allowance along with default settings; anything you imported as a guest lives only in your browser and is removed by a reset.
Creating an account
From the console account panel, choose Sign in or create account. On this documentation site, use Sign in in the top navigation. Both routes open the same hosted sign-in page where you can register with an email address, confirm it, and sign in. Corporate SSO is available for enterprise tenants.
Console account panel with the Sign in or create account action.
Signing in unlocks:
- A persistent identity across the console and docs, on every browser.
- A plan (Standard, Pro, or Enterprise) with its query allowance and ML tiers; see Plans and billing.
- The fleet integration surface described below: API tokens, the fleet agent handoff, and Live fleet mode.
- Persistent reports, saved scenarios, and chart library entries tied to your browser profile rather than a resettable guest session.
Signing out from either surface ends the hosted session for both.
Fleet integration
Signed-in accounts see Connect your fleet in the console: a four-step path from an empty console to live program telemetry. You can skip it at any point and return under Settings.
1. Create an API token
Tokens are created under Settings, in the API panel, with a name and scopes:
| Scope | What it covers |
|---|---|
| Topology | GET /topology (fleet graph and telemetry projection) |
| Predictions | GET /predictions?model_family=... (model outputs) |
| Telemetry | POST /telemetry (JSON batches, max 1,000 records) |
| Config | Reserved. Config writes are not exposed by the platform yet. |
Tokens are shown once at creation and can be revoked any time under Settings, then API. Two honest notes: scope enforcement is a platform roadmap item (today a platform key is valid for every exposed endpoint), and a token created in the console is a local credential until it is provisioned on the platform by the Constellation team; see API authentication for how platform keys work end to end.
You do not need Live fleet mode to send telemetry: the telemetry endpoint works with a token and the ingest URL alone, so you can ingest first and switch the timeline later.
2. Install the fleet agent
The Agent handoff panel generates tenant-specific artifacts embedding your token and
ingest URL: an agent bundle (JSON), IDE rules, a copyable install command, and a
config.toml. The telemetry routing mode controls how edge agents reach the API
(adaptive, direct, or console; console routing works in local development only). The
formats and the current agent caveats are documented in
Integration bundle.
3. Test the connection
The connection test probes the real platform, stage by stage, and reports each honestly instead of showing a fake success:
| Stage | What it checks |
|---|---|
| Platform API | GET /health reachability |
| Telemetry subsystem | GET /health/telemetry |
| Topology subsystem | GET /health/topology |
| Auth and data plane | Authenticated GET /topology (valid key, tenant routable, recent data) |
Every stage shows Not tested yet until you run it. The same staged probe is available as a script in Connection smoke test.
4. Open Live fleet
With telemetry arriving, switch the timeline to Live to replace the demo simulation with program telemetry. The globe, analytics, and the assistant all read the same live projection; Timeline and simulation covers how live mode relates to replay.
Your account portal
Signed-in users get a personal portal on this site at /account: identity and plan, plan upgrades through Stripe checkout, assistant usage meters with exact reset times, API tokens, saved reports (downloadable as Markdown), saved charts and scenarios, and assistant history, all in one page. Invoices, payment methods, and token management stay in the console; the portal deep-links to them.
The portal reads from the console running in the same browser over a secure, origin-checked channel, so it reflects the browser profile you use in the console. Artifacts created on other devices appear once the platform artifact store ships. If the portal cannot reach the console, open the console in this browser once and reload the page.
Sign-in remains optional for the documentation itself: every guide and reference page is public, and the portal is the only personalized page.
Where to go next
- Quick start for the ten-minute first session.
- Plans and billing for what each tier includes.
- Connect telemetry for the end-to-end ingestion tutorial.
- Permissions and tenancy for how accounts map to tenants and what enterprise isolation adds.