Quick Start

From signup to your first instrumented run in under 60 seconds.

1. Get an API key

2. Set the env var

bash

export INVARIANCE_API_KEY=inv_live_...

3. Instrument a run

Pick your stack — each block is the complete first run.

TypeScript

bash

npm install @invariance/sdk

instrument-run.tstypescript

import { Invariance } from '@invariance/sdk';

const inv = Invariance.init({ apiKey: process.env.INVARIANCE_API_KEY! });

await inv.runs.start({ name: 'hello-invariance' }, async (run) => {
  await run.log('prompt received', { prompt: 'Hello, world?' });
  await run.tool('greet', { who: 'world' }, async () => ({
    greeting: 'Hello, world!',
  }));
});

Python

bash

pip install invariance-sdk

instrument_run.pypython

from invariance import Invariance

inv = Invariance()  # reads INVARIANCE_API_KEY

with inv.runs.start(name="hello-invariance") as run:
    with run.step("greet", input={"who": "world"}) as s:
        s.output = {"greeting": "Hello, world!"}

CLI

bash

npm install -g @invariance/cli

bash

inv login --api-key "$INVARIANCE_API_KEY"
RUN=$(inv runs start --name hello --json | jq -r .id)
inv nodes write "$RUN" \
  --action-type tool_call \
  --input '{"who":"world"}' \
  --output '{"greeting":"Hello, world!"}'
inv runs update "$RUN" --status completed
inv runs inspect "$RUN" --json   # full run + nodes (agent-friendly)

MCP (for AI agents)

Add the MCP server to your Claude Desktop / Claude Code / Cursor config:

json

{
  "mcpServers": {
    "invariance": {
      "command": "npx",
      "args": ["-y", "@invariance/mcp"],
      "env": { "INVARIANCE_API_KEY": "inv_live_..." }
    }
  }
}

Then ask the agent: “Start a run named hello and write a node.”

4. View it in the dashboard

Open the Runs page in the dashboard — your run appears with its nodes, status, and a verifiable proof chain.

Next Steps

Concepts

Runs, nodes, monitors, signals, reviews.

CLI Reference

Full command reference for the invariance CLI.

Configuration

Env vars, API keys, SDK config.

Overview

Concepts

On this page

1. Get an API key 2. Set the env var 3. Instrument a run 4. View it in the dashboard Next Steps