db9 — Postgres but for agents

3 min read Original article ↗

Postgres but for agents_

Query it. Or cat it._

Create, manage, and query serverless PostgreSQL databases from your terminal. Branching, migrations, observability, type generation — all built in.

Full PostgreSQL and a cloud filesystem for AI agents. SQL when you need power. File ops when you need simplicity. One database, unified interfaces.

CLI

$ curl -fsSL https://db9.ai/install | sh

macOS / Linux (x86_64, arm64)

AI Agents

Read https://db9.ai/skill.md and follow instructions

Your agent learns to install, auth, and use db9 autonomously

DB9 works with any agent and stack

Claude Code OpenAI Codex Cursor Cline VS Code OpenCode OpenClaw Vercel m9mem9 Claude Code OpenAI Codex Cursor Cline VS Code OpenCode OpenClaw Vercel m9mem9

Personal assistants & customer copilots

Memory in tables. Context in files.

Structured state lives in Postgres. Raw context, transcripts, and session snapshots stay as files — all in one workspace.

Research, docs & coding agents

Knowledge from files. Retrieval from SQL.

Source documents stay in the filesystem. Chunks, metadata, and vectors live in Postgres — agents retrieve grounded context from one backend.

Automation, reports & multi-agent runs

Outputs in files. History in Postgres.

Reports, traces, and artifacts persist as files. Run history, status, and metadata live in Postgres — one workspace per agent run.

Everything your agent needs

Auto-embeddings, vector search, environment branching, file storage, cron — built in, not bolted on.

AI, built in

Auto-embeddings, vector search, and HTTP — all in SQL.

Call embedding() in a query and get vectors back — no external pipeline, no API keys in application code. Similarity search and outbound HTTP are native too.

replaces vector DB replaces embedding pipeline replaces HTTP glue

db9:myapp=>

-- similarity search with built-in embeddings
SELECT title, content
FROM docs
ORDER BY vec <-> embedding('deploy to production')
LIMIT 5;
-- generate embeddings inline
UPDATE docs SET vec = embedding(content)
WHERE vec IS NULL;
-- call external APIs from SQL
SELECT body::json>>'status'
FROM http_get('https://api.example.com/health');

Branch everything

Clone your entire environment. Not just tables.

One command creates an isolated copy — data, files, cron jobs, and user permissions. Test against real conditions, then delete it.

terminal

$ db9 branch create myapp --name staging
  Branch 'staging' created from database myapp.
  Name        staging
  State       ready
  Admin User  admin

Tables & rows Files & uploads Cron jobs User permissions Extensions

Built-in file storage

Upload, download, and mount files alongside your data. No S3 buckets to configure.

$ db9 fs cp ./data.csv myapp:/imports/
  data.csv -> /imports/data.csv

$ db9 fs mount myapp ~/local
  Mounted myapp at ~/local

Cron that never sleeps

Distributed job scheduling from SQL or CLI. No idle timeouts, no missed runs.

$ db9 db cron myapp create \
  '*/5 * * * *' \
  'SELECT sync_embeddings()'

  Cron job created.
  jobid: 1

Full Postgres. Fully typed.

Zero setup, 600+ ORM tests passing. One command to generate TypeScript or Python types.

$ db9 create
  Name  brave-tiger-42

$ db9 gen types brave-tiger-42
interface Document {
  id: number;
}

Start building in seconds

One command to install. One command to create a database. Zero config.

$ curl -fsSL https://db9.ai/install | sh

Native to your agent

Serverless Postgres for AI agents. Create, query, branch, and manage databases from the terminal — zero config. Built-in JSONB, vector search, HTTP extension, filesystem queries, cron jobs, and full-text search.

Read https://db9.ai/skill.md and follow the instructions

Unlock deep insights and enterprise-grade observability. Stream OpenClaw events into db9 and secure runtime events as immutable JSONL audit logs.

Read https://db9.ai/plugins/my-claw-dash.md and follow instructions to use my-claw-dash