AI-powered · Open source · Self-hostedby Aniket

Coinmaester

An AI-powered finance tracker for transaction emails. Sync your inbox, let a local model classify and extract spending details, and explore everything in a simple web UI.

How it works

Email in, structured transactions out — local AI does the classification on your host, not in Docker.

  1. 01

    Sync email

    Connect your inbox and pull transaction emails. Message bodies stay on disk; only paths live in the database.

  2. 02

    AI classify locally

    A host-run AI model (Hugging Face) labels messages and extracts amounts, merchants, and types — nothing leaves your machine for inference.

  3. 03

    Explore spending

    Browse transactions, analytics, and email sync status in the web UI once processing finishes.

Why local — and not a free online app?

Self-hosting is the product decision, not a missing feature.

  • Your data stays yours

    Transaction emails are sensitive. Coinmaester is designed so AI classification runs on your machine — you keep control instead of uploading finance mail to a third-party cloud for inference.

  • Built to run where you are

    An always-on hosted stack (API, database, queue, and especially AI inference) has real infrastructure cost. Rather than gate the product behind a paid SaaS, it ships as open source you can self-host today.

Built independently

Support the project

Coinmaester is built and maintained by Aniket Chanana. If you find it useful and want to collaborate, contribute ideas, or help make a future hosted option sustainable, reach out on any channel below.

Self-host in minutes

Run web and API in Docker; keep the Hugging Face worker on the host so it can use your CPU (or GPU).

terminal — self-host

Full OAuth URIs, ports, and networking notes live in the self-hosting guide.