Taiwan-built, self-hosted AI agent harness in a single Go binary — writes its own tools, sandbox-tests them, and shares them over MCP with Claude Code and Codex.
⚡ The Modern SQL Database Engine — Embedded or Standalone Server
MySQL Protocol · SQL + JSON · ACID · MVCC · Encryption · Replication · Pure Go
CobaltDB runs in two modes — use it as an embedded library inside your Go application, or deploy it as a standalone database server that any MySQL client can connect to.
Requirements: Go 1.25+ (toolchain go1.26.5) · Zero CGO runtime dependency
CobaltDB runs in two modes — use it as an embedded library inside your Go application, or deploy it as a standalone database server that any MySQL client can connect to.
Requirements: Go 1.25+ (toolchain go1.26.5) · Zero CGO runtime dependency
Verification and Security Checks
Run core checks locally:
make verify
Run full security/concurrency gate (requires CGO toolchain and gcc for race detector):
make verify-security
If tool commands are missing locally, install:
go install golang.org/x/vuln/cmd/govulncheck@latest
go install github.com/securego/gosec/v2/cmd/gosec@latest
go install github.com/golangci/golangci-lint/cmd/golangci-lint@latest
⚡ Quick Start
1. Server Mode — Connect with MySQL-Compatible Clients
# Start CobaltDB server
./cobaltdb-server -mysql-addr 127.0.0.1:3307 -admin-pass "StrongPass123!" -data ./data
# Connect with standard MySQL client
mysql -h 127.0.0.1 -P 3307 -u admin -p
mysql> CREATE TABLE users (id INTEGER PRIMARY KEY, name TEXT, email TEXT);
mysql> INSERT INTO users VALUES (1, 'Alice', '[email protected]');
mysql> SELECT * FROM users;
+----+-------+-------------------+
| id | name | email |
+----+-------+-------------------+
| 1 | Alice | [email protected] |
+----+-------+-------------------+
The MySQL wire-protocol surface is tested in-repo for handshake, COM_QUERY,
result sets, and parameterized prepared statement execution. Validate each
production driver or ORM against docs/MYSQL_COMPATIBILITY.md.
Server features: MySQL protocol, TLS 1.2+, authentication (Argon2id), connection pooling, rate limiting, circuit breaker, health checks, encrypted tamper-evident audit logging, and master-slave replication transport.
Production note: CobaltDB is a production-oriented single-node database candidate. Use TLS for non-loopback wire access, keep the MySQL listener private or disabled, and do not treat replication as automatic HA/failover infrastructure. Replication plaintext is now limited to syntactic loopback endpoints for local development; non-loopback replication requires authenticated TLS, with optional mutual TLS and SSLServerName certificate-name override on replicas.
2. Embedded Mode — Use as a Go Library
import "github.com/cobaltdb/cobaltdb/pkg/engine"
db, _ := engine.Open(":memory:", &engine.Options{InMemory: true})
defer db.Close()
ctx := context.Background()
db.Exec(ctx, `CREATE TABLE users (
id INTEGER PRIMARY KEY,
name TEXT NOT NULL,
email TEXT UNIQUE,
metadata JSON
)`)
db.Exec(ctx, "INSERT INTO users VALUES (?, ?, ?, ?)",
1, "Alice", "[email protected]", `{"role": "admin"}`)
rows, _ := db.Query(ctx, `SELECT name, JSON_EXTRACT(metadata, '$.role') FROM users`)
Zero CGO, zero dependencies. Import and use — no external server needed.
CLI Mode
# Interactive shell
./cobaltdb-cli -path ./mydb.db
# Execute SQL directly
./cobaltdb-cli -memory "SELECT * FROM users"
# Server mode uses the MySQL protocol
mysql -h 127.0.0.1 -P 3307 -u admin -p
Docker Mode
# Start with Docker Compose (includes Prometheus + Grafana monitoring)
export COBALTDB_ADMIN_PASSWORD='change-this-before-production'
export GRAFANA_ADMIN_PASSWORD='change-this-too'
docker compose up -d
# Or run standalone
docker build -t cobaltdb .
docker run -d \
-p 4200:4200 \
-p 3307:3307 \
-p 8420:8420 \
-e COBALTDB_ADMIN_PASSWORD='change-this-before-production' \
-e COBALTDB_ALLOW_CLEARTEXT_AUTH=true \
-v cobaltdb_data:/data/cobaltdb \
cobaltdb
# Connect to containerized database
mysql -h 127.0.0.1 -P 3307 -u admin -p
See DOCKER.md for detailed Docker setup instructions.
🌍 Multi-Language SDKs
CobaltDB speaks a tested subset of the MySQL wire protocol. Use standard MySQL
drivers for validated workflows, and run the compatibility checks for each
production driver or ORM.
-- DDL: Schema Definition
CREATE TABLE products (
id INTEGER PRIMARY KEY,
name TEXT NOT NULL,
price REAL CHECK (price > 0),
category TEXT,
tags JSON,
created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);
CREATE INDEX idx_category ON products(category);
CREATE VIEW expensive_products AS
SELECT * FROM products WHERE price > 100;
-- DML: Data Manipulation
INSERT INTO products (name, price, category, tags)
VALUES ('MacBook Pro', 1999.99, 'Electronics', '["laptop", "apple"]');
-- Complex SELECT with JOINs, Aggregates, Window Functions
SELECT
p.category,
COUNT(*) as total,
AVG(p.price) as avg_price,
MAX(p.price) as max_price,
ROW_NUMBER() OVER (PARTITION BY p.category ORDER BY p.price DESC) as rank
FROM products p
LEFT JOIN orders o ON p.id = o.product_id
WHERE p.price BETWEEN 100 AND 500
GROUP BY p.category
HAVING COUNT(*) > 5
ORDER BY avg_price DESC
LIMIT 10;
📊 Advanced Features
Window Functions
SELECT
name,
salary,
AVG(salary) OVER (PARTITION BY dept) as dept_avg,
RANK() OVER (ORDER BY salary DESC) as salary_rank,
LAG(salary) OVER (ORDER BY salary) as prev_salary
FROM employees;
JSON Operations
-- Extract nested values
SELECT JSON_EXTRACT(metadata, '$.user.address.city') FROM users;
-- Modify JSON
UPDATE users SET metadata = JSON_SET(metadata, '$.last_login', '2026-03-02');
-- Array operations
SELECT * FROM products WHERE JSON_ARRAY_LENGTH(tags) > 2;
Transactions (ACID)
BEGIN;
UPDATE accounts SET balance = balance - 100 WHERE id = 1;
UPDATE accounts SET balance = balance + 100 WHERE id = 2;
-- Atomic transfer
COMMIT;
-- Or ROLLBACK on error
Triggers & Procedures
-- Audit logging trigger
CREATE TRIGGER audit_log
AFTER INSERT ON users
BEGIN
INSERT INTO audit (table_name, action, record_id, created_at)
VALUES ('users', 'INSERT', NEW.id, CURRENT_TIMESTAMP);
END;
-- Stored procedure
CREATE PROCEDURE transfer_funds(from_id INT, to_id INT, amount REAL)
BEGIN
UPDATE accounts SET balance = balance - amount WHERE id = from_id;
UPDATE accounts SET balance = balance + amount WHERE id = to_id;
END;
CALL transfer_funds(1, 2, 100.00);
-- DDL
CREATE TABLE ... [PRIMARY KEY] [NOT NULL] [UNIQUE] [CHECK] [FOREIGN KEY]
CREATE INDEX ... ON ...
CREATE VIEW ... AS SELECT ...
CREATE TRIGGER ... BEFORE|AFTER INSERT|UPDATE|DELETE
CREATE PROCEDURE ...
DROP TABLE|INDEX|VIEW|TRIGGER|PROCEDURE ... [IF EXISTS]
-- DML
INSERT INTO ... VALUES (...)
INSERT INTO ... SELECT ...
UPDATE ... SET ... WHERE ...
DELETE FROM ... WHERE ...
SELECT ... FROM ... [JOIN ... ON ...] [WHERE ...] [GROUP BY ...] [HAVING ...] [ORDER BY ...] [LIMIT ...] [OFFSET ...]
-- DCL
BEGIN | COMMIT | ROLLBACK
CALL procedure_name(...)
🔧 Development
# Clone the repository
git clone https://github.com/cobaltdb/cobaltdb.git
cd cobaltdb
# Run tests
go test ./... -v
# Run tests with coverage
go test -coverprofile=coverage.out ./...
go tool cover -func=coverage.out
# Run benchmarks
go test -bench=. -benchtime=2s ./test/...
# Build CLI
go build -o cobaltdb-cli ./cmd/cobaltdb-cli
# Build Server
go build -o cobaltdb-server ./cmd/cobaltdb-server
# Run demo
go run cmd/demo/main.go
Test Coverage
Package
Coverage
Package
Coverage
pkg/logger
100.0% ✅
pkg/pool
97.5% ✅
pkg/advisor
96.8% ✅
pkg/auth
96.8% ✅
pkg/wire
94.7% ✅
pkg/metrics
94.3% ✅
pkg/fdw
93.9% ✅
pkg/optimizer
93.8% ✅
pkg/security
93.2% ✅
pkg/query
91.7% ✅
pkg/parallel
91.2% ✅
pkg/cache
90.9% ✅
pkg/audit
90.7% ✅
pkg/server
90.6% ✅
pkg/storage
88.4% ⚠️
pkg/replication
88.3% ⚠️
pkg/backup
88.2% ⚠️
pkg/txn
87.8% ⚠️
pkg/catalog
84.8% ⚠️
pkg/wasm
84.4% ⚠️
pkg/protocol
82.7% ⚠️
pkg/engine
80.8% ⚠️
pkg/btree
78.2% ⚠️
7,100+ test functions across 600+ test files, all passing. Current pkg/... coverage is 86.3%; several production-critical packages remain below the 90% package target.
