get started

Start small. See it work.

Try a working dashboard with synthetic invoices, a mock model and a report your agent can read. A useful first look, on your own machine.
Start an evaluationInstructions for your agent
your first look

Two commands, once you have access.

  1. Get repository access

    The product repository is private. Email us to arrange evaluation access and confirm the terms.

  2. Run it locally

    Use Node.js 24 and run these commands from your authorized product checkout. Installing dependencies needs registry access.

  3. Follow a charge. Inspect a trace.

    Open the printed dashboard URL. Sign-in details are in the private file the command names. Explore the sample subscriptions, invoices and recorded request.

product repository · Node.js 24
npm ci
npm run evaluate

Synthetic data. Mock provider. No live AI keys.

Ctrl+C stops the evaluation and removes its temporary data.

Evaluation terms

It checks the work as it goes.

These are local application checks using synthetic data. Live integrations and production deployment need their own verification.

  • Invoice import and duplicate protection
  • Approved and prohibited subscription register
  • Billed API and subscription amounts kept separate from estimates
  • Mock gateway request with captured conversation evidence
  • MCP reporting scoped to Engineering; Finance evidence excluded
bring your coding agent

Give it a clear starting point.

Paste this into your coding agent once you have an authorized product checkout. The full guide includes prerequisites, checks and what to do if a step fails.

agent prompt

Read https://tokenobserve.com/agent-setup.txt and help me run the local Token Observe evaluation from my authorized product checkout. Check the prerequisites, run the automated evaluation and report which checks passed. Keep credentials private and report any blocker.
Read the full agent guideDownload .txt

For a check that exits and cleans up

npm ci
npm run evaluate -- --check
next, one of your agents

Give it an identity. Connect your client.

On your own installation, an operator configures the provider, registers the agent, assigns its permissions and budget, then issues its key. Your supported client can use that key and the gateway address.

Replace the example host and key with the values from your operator. Check your client version against the integration guide.

OpenAI-compatible client

Use the gateway address with /v1 and the key issued for your registered agent.

OPENAI_BASE_URL="https://YOUR-PRIVATE-HOST/v1"
OPENAI_API_KEY="<issued-agent-key>"
OpenAI integration guide
Anthropic client

Use the gateway address without /v1 and the key issued for your registered agent.

ANTHROPIC_BASE_URL="https://YOUR-PRIVATE-HOST"
ANTHROPIC_API_KEY="<issued-agent-key>"
Anthropic integration guide

Want your agent to read the AI bill?

Use your installation’s /mcp endpoint with an operator-issued agent or seat bearer and explicit report and team permissions. The agent guide includes the connection sequence and a spend-report call.

MCP connection instructionsDeployment requirements and limits
your next step

Let’s get you a first look.

Email hello@tenhaw.com for repository access. Tell us what you’d like to understand about your AI spend or activity.

Start an evaluationRead the evaluation terms