We’ve been developing the Tonk stack & CLI for the last 3 months alongside indie hackers and design partners. The tool is currently in alpha, but powerful enough to build useful tools. Tonk is a novel approach to using AI, so we wrote this explainer as a primer for new users.
Tonk is a stack and CLI designed to simplify the process of building custom dashboards, tools, and AI agents using your own data.
It's aimed at "vibe coders" - semi-technical people who want to quickly build personalised tools and automations - without backend headaches.
The problem Tonk solves
Creating useful personal tools means you have to:
Pull in diverse data (your notes, emails, files, or external APIs & scraped web data)
Manage backend tasks (data storage, synchronisation, security, offline access)
Traditionally, these tasks require significant effort, making quick experimentation difficult. But in our experience, the joy of vibe coding is strongest when you’re freestyling in a seamless cybernetic loop with your data.
How Tonk works
The Tonk CLI includes a list of commands that make it easy to:
Set up "Workers": Services that automatically pull data into your personal "Store".
Manage a "Store": Your private, local-first data repository that integrates various data sources.
Build apps, services & agents: Lightweight, custom tools that use this data.
We’ve been building with AI for the last 12 months and in our experience, this structure is by far the best for simplifying experimentation - allowing you to rapidly iterate between ingesting new data and creating or refining apps.
This flow unbundles data structures from ingestor and application logic for maximum flexibility. It’s analogous to a data lake, but for individuals and small teams.
Tonk is not a vibe coding agent like Cursor, Claude Code or Windsurf. Tonk sits around these as a streamlined stack.
Example workflow
Using a few Tonk CLI commands:
You create and run a Tonk Worker to pipe your calendar events and emails into your personal store.
You quickly build an agent that summarizes your daily agenda and emails every morning.
Realising you also want task integration, you set up another worker to pipe in your task management data.
You enhance your agent to combine tasks with your schedule, automatically generating prioritized daily to-dos.
What you can build with Tonk
Dashboards: Financial tracking, CRM, personal productivity tools
AI Agents: Summarizers, coaches, personalized recommendation engines
Custom Tools: Data visualizers, task managers, note-sharing services
Additional benefits
Local-first: Data is stored locally by default, ensuring privacy and offline availability.
Sync engine: Tonk apps natively support collaboration with multi-user real-time updates.
No database complexity: No migrations or caching headaches - and all state management is taken care of.
Early GUI preview: introducing TonkbookLM
For those who prefer a graphical interface, TonkbookLM provides an approachable version of Tonk’s best features, focused on personal knowledge management. It's similar in concept to Google's NotebookLM but with much deeper extensibility.
Watch Jack introduce TonkbookLM:
(TonkbookLM is currently in pre-alpha and requires some manual setup to use. See the repo here)
Limitations and next steps
Tonk is currently in alpha. Expect rough edges, limited documentation, and ongoing development. We actively work with indie hackers experimenting with the tool, so if you’re using it please reach out to the team.
Upcoming features include:
Easier user space and security setup
Easier hosting and deployment options
A decentralised data-sharing network for collaboration across Tonk stores - for friends, communities and enterprise teams
Getting started
Check out the Tonk documentation for quickstart instructions and installation details.
Reach out to the Tonk Team on either Whatsapp or Telegram for support.
Happy hacking!
FAQ
I heard Tonk was an applied cryptography startup. What’s the link?
Our goal is to make it easy to connect various Tonk stores to maximise data interoperability - in a flexible, private, self-sovereign, peer-to-peer manner so that it’s impossible for anyone to build a corporate walled garden. This will require a novel data sharing protocol using advanced cryptography such as ZK and MCP.
Tonk connects data together. How is this different to MCP?
MCP is a way to feed context and external tools into an LLM. Tonk provides a novel stack that enables LLMs to only write React code, while interfacing with a complex system of data persistence, real-time sync, offline capabilities, and more. While MCP helps make vibe coding more context-aware and capable, Tonk streamlines the experience of vibe coding over your data.
What’s the vision?
We want to break the 2010s economics of the internet so that anyone can do anything with their data - living through their own software and collaborating over their own networks. We see AI as a monumenal force that massively increases the value of self-sovereign data - not as something you trade off against convenience, but a necessary precondition of amazing user experiences.