Ghostgres

Are you tired of maintaining your relational database? Tired of having to write "syntactically correct" SQL queries? Tired of having queries fail because tables "don't exist"?

Look no further. Ghostgres is the database of the future. There's nothing to maintain. No query language. It doesn't return errors. The only limits on what you can query are the limits of your imagination.

The best part? It's Postgres wire-compatible, so you can plug it in wherever you use Postgres.

Just grab an API key for your favorite model provider and connect:

psql "postgres://openai:<OPENAI_API_KEY>@try.ghostgres.com/gpt-5.4"

Then start querying:

gpt-5.4=> What is the best database?;

Just remember to end your queries with a semicolon (;).

Oh, and be careful with single (') and double quotes (").

Built with <3 by Ghost.

Connecting

Connect using psql or any Postgres client. The username is the LLM provider (openai or anthropic), the password is your API key, and the database is the model name.

psql "postgres://<provider>:<api_key>@try.ghostgres.com/<model>"

Running Locally

You can run the Ghostgres server locally with:

go run github.com/timescale/ghostgres/cmd/ghostgres@latest

Then connect to localhost:

psql "postgres://openai:<OPENAI_API_KEY>@localhost/gpt-5.4"

Flags:

The server accepts the following flags:

Flag	Default	Description
-host	"" (all interfaces)	Hostname/interface to bind to
-port	5432	Port to listen on
-log-level	info	Log level (debug, info, warn, error)
-prompt	Built-in prompt	Path to a file containing a custom system prompt
-tls-cert	(disabled)	Path to TLS certificate PEM file (requires -tls-key)
-tls-key	(disabled)	Path to TLS private key PEM file (requires -tls-cert)

Supported providers

OpenAI

Username: openai. Password: your OpenAI API key. Database: any OpenAI model name (e.g. gpt-5.4, gpt-4o, o3).

Options:

Option	Description	Default
reasoning_effort	Reasoning effort level (e.g. none, minimal, low, medium, high). Valid values depend on the model.	Lowest supported for the model

Anthropic

Username: anthropic. Password: your Anthropic API key. Database: any Anthropic model name (e.g. claude-sonnet-4-6, claude-opus-4-6).

Options:

Option	Description	Default
effort	Output effort level (low, medium, high, max)	low
thinking	Extended thinking budget in tokens (minimum 1024)	Disabled

Setting options

Options are passed via the options connection parameter (PGOPTIONS env var or options= in the connection string). Multiple options are space-separated.

psql "postgres://openai:sk-...@localhost/gpt-5.4?options=reasoning_effort%3Dhigh"

# Or:

PGOPTIONS="reasoning_effort=high" psql "postgres://openai:sk-...@localhost/gpt-5.4"

How it works

Each client connection gets its own LLM chat session. The server translates Postgres wire protocol messages into LLM API calls and converts the structured JSON responses back into proper Postgres result sets. Chat history is maintained per connection, so the LLM can stay consistent across queries within a session.

SSL/TLS

Ghostgres supports optional SSL/TLS encryption via the PostgreSQL wire protocol's SSLRequest handshake. When enabled, clients that request SSL (e.g. sslmode=require) will have their connections upgraded to TLS before any credentials are sent.

To enable TLS, provide a certificate and private key:

ghostgres -tls-cert /path/to/cert.pem -tls-key /path/to/key.pem

When TLS is not enabled, SSL requests from clients are denied and the connection continues in plaintext. Clients using sslmode=prefer (the default for most Postgres clients) will fall back to an unencrypted connection automatically.

Security note

Authentication uses cleartext passwords because the server needs the raw API key to call the LLM provider. When running remotely, enable TLS (see above) to encrypt the connection. Without TLS, connect over localhost or use SSH tunneling.