Start with one billing source.
Create. Connect. Check the source.
Create and confirm your account
Open the account form and enter your email and a password of at least 12 characters. Follow the confirmation link in your email, then sign in.
Create a workspace
The service provisions the workspace. Its free 30-day trial begins at creation, with no card and no Token Observe server installation.
Connect a billing source
Supply the required reporting credential for OpenAI, Anthropic Claude, xAI Grok or Google Cloud Billing. Check currencies, reporting periods, last collection time and partial or failure status.
Billing permissions and Google export setup may require work in your provider's console. Saving a separate inference key does not send a model request.
After 30 days, workspace changes, model calls and scheduled collection stop. Authorized retained reads and exports remain subject to the shared traffic quota.
Trial scope and limitsPrepare your provider reporting access
Read the evidence with its limits.
Billing connections do not recover historical prompts or browser conversations. Provider model calls, model-assisted search and Google BigQuery scans can incur charges. This is a small cloud pilot without an availability SLA.
- Reported provider costs, currencies and source periods
- Collection freshness and partial or failed source status
- Gateway estimates kept separate from reported costs
- Activity only for requests routed through the workspace or configured sources
- Metadata capture by default; redacted text capture is opt-in
Give it a clear starting point.
Use this prompt for help with the cloud journey and supported client configuration. Complete account confirmation yourself and keep credentials out of the chat.
agent prompt
Read https://tokenobserve.com/agent-setup.txt and help me set up a Token Observe cloud workspace. Explain the required billing permissions, distinguish reporting credentials from inference and workspace keys, and help configure a supported client. Keep secrets private and report actual blockers.Issue an identity. Connect your client.
Supply a separate inference key and model in your workspace. Register the agent, assign permissions and a budget, then issue its scoped key. Configure your supported client with that key and the workspace gateway address.
Use the exact gateway address shown for your workspace and check your client version against the integration guide. Signup alone does not issue a machine key.
OpenAI-compatible client
Use the workspace gateway address ending in /v1 and the key issued for your registered agent. Check the request shapes your client supports.
OPENAI_BASE_URL="<workspace-gateway-base-url>"
OPENAI_API_KEY="<issued-workspace-agent-key>"Anthropic client
Use the workspace gateway base address supplied by the portal and a workspace-issued agent key. Native requests reach /w/<workspace-id>/v1/messages.
ANTHROPIC_BASE_URL="<workspace-gateway-base-url>"
ANTHROPIC_API_KEY="<issued-workspace-agent-key>"Want your agent to read the AI bill?
Use the hosted /w/<workspace-id>/mcp endpoint with an active workspace-issued agent or seat key and explicit report and team grants. A provider key or human sign-in token cannot authenticate MCP. The agent guide describes discovery and read-only access.
Start your first cloud workspace.
Create your account, confirm your email and connect a supported billing source. The 30-day trial begins at workspace creation, with no card.