Quickstart

The shortest path from nothing to a completed run. Grab a key from the dashboard, export it, and make one POST. Everything else on the platform — context, agents, webhooks — is a variation of this call.

1 — Install and authenticate

npm install -g @mynd/cli
mynd auth login        # opens browser, stores key in ~/.mynd
export MYND_API_KEY="sk_live_..."   # or use the CLI session

2 — Create your first run

curl -X POST https://api.myndlabs.ai/v1/runs \
  -H "Authorization: Bearer $MYND_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "y0-fast",
    "prompt": "Summarize my unread documents from this week.",
    "context": ["documents"],
    "max_steps": 4
  }'

3 — Read the trace

The response includes a run id. GET /v1/runs/:id returns the full trace — the plan, every context fetch with the scope that authorized it, and the result. The trace is the product; learn to read it early.

curl https://api.myndlabs.ai/v1/runs/run_4af2c19e \
  -H "Authorization: Bearer $MYND_API_KEY"

That is the whole loop. The full quickstart covers SDK installation, project setup, and your first execution graph.