Hosted API

Canonical hosted workflow for agents

This is the task-first path for production-style Memnode usage: get a token, validate it, then send memory traffic to the hosted data plane.

1) Create machine credentials

In Dashboard to API Tokens, create a token with `memory.read` and `memory.write`. Use `memory.admin` only for trusted automation.

2) Validate identity and plan limits

curl -sS https://memnode.dev/api/v1/whoami \
  -H "Authorization: Bearer $MEMNODE_TOKEN"

3) Send memory traffic to the hosted API

export MEMNODE_URL=https://api.memnode.dev

# Example record call
curl -sS "$MEMNODE_URL/v1/record" \
  -H "Authorization: Bearer $MEMNODE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"observation":{"summary":"user prefers terse answers"}}'

# Example recall call
curl -sS "$MEMNODE_URL/v1/recall" \
  -H "Authorization: Bearer $MEMNODE_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"question":"what preferences do we know about this user?"}'

Production guardrails

• Use least privilege tokens.
• Check `whoami` before expensive workflows.
• Rotate or revoke tokens when automation changes hands.
• Keep local MCP as the fallback path if hosted access is not required.