Getting started

Install pk, wire it to your AI tool, and start a session.

Prerequisites

Git tracks knowledge operations

Bun runtime and package manager

The curl installer handles all of these automatically.

1. Install

install + init

$curl -fsSL https://justestif.github.io/pk/install.sh | bash

$pk init

Run pk init in your project folder. It creates a knowledge store at ~/.pk/<project>/ and wires up your AI tool automatically.

2. Tell your agent what you know

Open your AI tool and tell it about your project. Your agent logs it — you never touch the CLI directly.

You log that we decided to use Postgres as the primary database

Agent Saved to decisions/use-postgres.md

Log decisions you've already made, open questions, and constraints. The more context your agent has on day one, the less time it spends re-deriving what you already know.

3. Start a session

Open your AI tool and type "use pk skill". This loads the playbook that tells your agent how to search, log, and update notes.

You use pk skill

Agent Loaded. I can now search your notes, log decisions, track open questions...

From then on, your harness injects a session summary automatically — you don't need to ask again.

4. Set up embeddings

recommended

Without embeddings, search only matches exact keywords — searching database latency won't find a note titled "slow queries." With embeddings, pk uses hybrid search automatically. Set it up once.

embeddings setup

# install Ollama from ollama.com, then:

$ollama pull nomic-embed-text

$pk config --embedding nomic-embed-text

$pk index

Runs locally — no GPU required, ~274 MB download. Full setup guide →

New to AI coding tools?

The full guide walks through installing Git, Bun, your AI tool, and pk step by step.

Full guide →

Something broken or unclear? Open a GitHub issue