Administer an account
Admin surfaces manage service access, plan and usage, LLM credentials, organization identity, and compliance evidence. Organization owners and admins can see more controls than ordinary members.
Access
Access inventories managed service state:
- Langfuse — observability workspace and login state
- LiteLLM — model-routing access and connection state
- Gitea organizations — source workspace identity and links
- Repositories — agent repository access and managed credential references
Copy values only into the client that needs them. Do not put managed service credentials in an agent card, source repository, prompt, or browser-visible frontend configuration.
Billing
Billing has:
- Overview — current plan and account posture
- Plan — upgrade, portal, and cancellation controls
- Usage — account and per-agent consumption
- Invoices — provider invoice history
- Credits — prepaid balance and top-ups for paid agent calls
- Payouts — creator account, earned/paid/owed totals, and settlements
- Feature gates — account features enabled by the current plan or rollout
Subscription invoices, prepaid call credits, creator payouts, and runtime receipts are related but distinct ledgers.
LLM keys
LLM Keys lets an account:
- add an OpenAI-compatible provider key and base URL
- choose a model from the provider-compatible catalog
- select the default credential
- replace or remove saved credentials
- inspect model health
Agents using ctx.llm receive a short-lived platform route, not the provider
secret in public metadata. See LLM credentials.
Organizations
Organization covers:
- verified domains
- SCIM configuration and bearer tokens
- members and owner/admin/member roles
- activation state
- audit history
SCIM tokens are shown only when created. Store them in the identity provider, rotate them deliberately, and use the organization audit log to verify member changes.
Compliance
Compliance provides:
- Decision Records — tamper-evident governance decisions
- Policy — retention and organization compliance settings
- Agent Risk — classification and review state per agent
- Evidence export — organization-scoped evidence packages
The dashboard helps assemble EU AI Act readiness evidence; it does not replace legal classification or an organization's accountable human decision.
Recommended operating cadence
- Daily: review failed work, spend posture, approval queues, and setup health.
- Before a release: inspect deployment, card, proof, evidence, and policy.
- Monthly: reconcile usage, invoices, credits, creator payouts, members, and audit events.
- On personnel or provider change: rotate SCIM, service, imported-agent, and LLM credentials.
See the Control-plane API guide for automation and the complete API inventory for every route.