AI coding agents stall the moment they hit "I need a database." instanode.dev removes that stall. One HTTP call, one postgres:// URL back, agent keeps going.
curl -s -X POST https://api.instanode.dev/db/new -H 'Content-Type: application/json' -d '{"name":"demo"}' | jq and use the connection_url from the response."
When an agent is writing code with you and needs a database, it does one of three bad things:
docker run postgres — fine if you have Docker, broken otherwise, and permanent context clutter either way.process.env.DATABASE_URL and hopes you'll figure it out later — the code doesn't run.None of these are the agent's fault. They're the fault of infrastructure tools that assume a human is in the loop.
A Postgres provisioner that speaks HTTP, requires zero authentication for the first call, and returns a real, working connection URL. Free tier is 24-hour disposable. Paid tier is $12/month and the database persists forever.
pgvector is pre-installed on every database. Connection URLs are standard postgres://user:pass@pg.instanode.dev:5432/db?sslmode=require — any driver works, no proprietary client.
CLAUDE.md / .cursorrules so agents learn the pattern and propagate it to future projects.claude mcp add instanode -- npx -y @instanode/mcp@latest (npm · source)pip install instanode (PyPI)pip install langchain-instanode (PyPI)pip install llama-index-tools-instanode (PyPI)pip install crewai-instanode-tools (PyPI)
./scripts/setup-db.sh, pnpm dev.