Getting started

Requirements

• Go 1.26 or later
• Any supported operating system: Linux, macOS, BSD or Windows. AbyssDB uses the storage device built into the operating system, so there is nothing else to install, provision or size.

Build

git clone https://github.com/noctarius/abyssdb
cd abyssdb
go build -o abyssdb ./cmd/abyssdb
./abyssdb version

Your first session

Start the interactive shell:

$ ./abyssdb
AbyssDB 0.0.0-null (codename: "Event Horizon")
storage engine: Abyss kernel-native (POSIX)
Type SQL terminated by ';'. \? for help, \q to quit.

abyss=>

Statements end with ; and may span multiple lines. Create a table, write to it and read it back:

abyss=> CREATE TABLE customers (id BIGINT PRIMARY KEY, name TEXT NOT NULL);
CREATE TABLE
Time: 0.000 ms

abyss=> INSERT INTO customers (id, name) VALUES (1, 'Acme'), (2, 'Globex');
NOTICE:  2 row(s) written to customers
INSERT 0 2
Time: 0.000 ms

abyss=> SELECT id, name FROM customers ORDER BY name;
 id | name
----+------
(0 rows)
Time: 0.000 ms

Every result is exact. There is no eventual consistency to wait for and no cache to warm up.

Useful shell commands

Run a single statement

./abyssdb exec -e "SELECT 1;"
echo "SHOW STATS;" | ./abyssdb exec

Connect with psql

Start the server, then connect with any PostgreSQL client:

./abyssdb serve
psql "postgresql://alice@localhost:5432/abyss?sslmode=disable"

NOTICE:  welcome to AbyssDB, alice. Nobody can access your data.
abyss=>

For encrypted connections with post-quantum key exchange, start the server with --tls and connect with sslmode=require. See PostgreSQL compatibility.

Connect an AI agent

claude mcp add abyssdb -- /path/to/abyssdb mcp

See AI integration for Claude Desktop, Cursor and HTTP setups.

Scale out

Run a sharded, replicated cluster in-process:

./abyssdb --shards 4 --nodes 3

Each of the 4 shards is a 3-node NullRaft group. Multi-shard transactions use two-phase commit automatically. See Architecture and Operations.

Next steps

• SQL reference
• Security: encryption is already on; learn how it works
• Operations: monitoring, chaos testing and benchmarking