Five ways to connect an agent

Connect an AI agent with the Python or Node.js SDK, the HTTP API, OpenTelemetry, or Toolcaise Connect for enforced limits, Claude Code and Codex.

Start free Open the connection guide

.py

Python

Use the Python SDK to report a run lifecycle and meaningful steps. Add checkpoints and guarded actions when you want compatible human control.

SDK reference →

.js

Node.js

Use the Node.js SDK from the process that owns the work. Report outcomes and usage where available, then inspect them in the workspace.

SDK reference →

POST

HTTP contract

Connect software that can make outbound HTTPS requests. Implement the documented reporting contract and use an agent-scoped credential.

API reference →

OTLP

OpenTelemetry

Point your trace exporter at Toolcaise with an agent credential: three environment variables, no Toolcaise package. Runs, model and tool spans and token counts arrive from your framework’s own tracing. It observes only; there is no pause or approval over OpenTelemetry.

Setup in the connection guide →

CLI

Toolcaise Connect

Run an agent under toolcaise-connect run for enforced cost, token and runtime limits and a stop from the dashboard. Put the MCP tool gateway in front of MCP servers, or turn on the Claude Code and Codex adapters to report their telemetry. A preview; builds are not yet code-signed.

Connect CLI reference →

Start by reporting runs, then add control where it matters

Report a heartbeat, an actual run, and a useful outcome. A standalone heartbeat confirms contact but does not instrument the rest of your workflow. Docker and cloud workloads can use the same SDK or HTTP paths inside the environment you already operate.

No model-provider key or inbound port is required. Keep agent credentials in secure runner configuration, and send only the operational metadata needed for supervision.

What each path can see, and what it can stop

Every path shows you the work. Only a Connect supervised run or the MCP tool gateway sits on the agent’s path and can stop it.

Cooperative

Python and Node.js SDKs, HTTP

What Toolcaise sees: What the agent reports: runs, steps, usage, outcomes
What it can stop or hold: Pause, resume, cancel and exact-action approvals, at checkpoints your code implements. Cooperative: code that skips the checkpoint is not stopped.

Observes only

OpenTelemetry

What Toolcaise sees: Runs, model and tool spans, and token counts from your framework’s own tracing
What it can stop or hold: Nothing. It observes only.

Observes only

Connect: Claude Code and Codex adapters

What Toolcaise sees: The allow-listed events and metrics those tools emit
What it can stop or hold: Nothing. It reports telemetry only.

Enforced on your machine

Connect: supervised run

What Toolcaise sees: Every model call made through an OpenAI- or Anthropic-compatible API, measured on your machine, plus the process’s start, end and resources
What it can stop or hold: Runtime, cost, token and model-call limits and a model allowlist, by stopping the whole process tree. Pause, resume, stop and limits from the dashboard.

Enforced at the proxy

Connect: MCP tool gateway

What Toolcaise sees: Every tool call that passes through the proxy
What it can stop or hold: Allow or deny by tool name, a call budget per session, and a hold until a person approves the exact call (Connect 0.3 or later). The approver sees the tool name; the arguments stay on your machine.

What no connection can do

Tell whether an agent’s work is correct. A person reviews that.
Stop an agent that goes around every control point.
Measure model calls sent to an endpoint Connect doesn’t sit in front of. The runtime limit and the stop still end the run.

Compare the ways of connecting →

Framework compatibility depends on your integration

An agent built with a framework can report through these paths when you can instrument its code, export its traces, or run it under Connect. This is not a claim of a native connector to every agent framework. SDK controls need compatible checkpoints; monitoring alone does not add them.

Framework quickstarts →Download the Toolcaise Connect preview →

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.

Start free Talk to us