Instrument an existing script, cloud worker, or workflow with the Python or Node.js SDK, the HTTP contract, or the OpenTelemetry traces it already emits, or run it under Toolcaise Connect.
No model-provider key, inbound port, or replacement agent runtime is required. You choose which operational metadata to report.
Monitoring does not automatically confer control. Add compatible checkpoints and action guards wherever you need a human decision, or run the agent under Toolcaise Connect to enforce cost, token and runtime limits.
Frequently asked questions
What is Toolcaise?
Toolcaise is a supervision workspace for AI agents you already run. It records each run and what it cost, opens an incident when scheduled work goes missing or stalls, lets a person accept or reject what an agent produced, and holds consequential actions until someone approves them. With Toolcaise Connect it can also stop a run at a cost, token or runtime limit.
Do I have to move my agents into Toolcaise?
No. Your agents keep running in your own infrastructure and report over outbound HTTPS through an SDK, the HTTP API, OpenTelemetry or Toolcaise Connect. Toolcaise never runs your code and never needs your model-provider keys.
How long does it take to connect the first agent?
For an agent you can edit, adding the SDK and wrapping the main loop in a run is a short change. For a command you cannot edit, Toolcaise Connect wraps it without code changes. The setup screen confirms the first report as soon as it arrives.
Which parts can actually stop an agent?
A run wrapped by Toolcaise Connect is stopped, with every process it started, at the limits you set or from the dashboard. Through the SDKs, pause, cancel and approvals are cooperative and apply at checkpoints in your code. OpenTelemetry only observes. The integrations page lists exactly what each path can enforce.
Connect your first agent
Create a free workspace, connect the agent you already run, and watch its next run arrive. Free covers 3 agents, with no card required.