About

Why this exists

I got tired of paying monthly subscriptions for tools that almost do what I need. Cloud storage that's too expensive at scale. Chat interfaces locked to one model. Forums I can't query with an API. Context management that doesn't understand my workflow.

Tokenburner is the answer: build exactly the features I need, running in my own AWS account, paying only for what I use. No seat licenses. No tier upgrades. No vendor deciding to sunset the feature I depend on.

Why Bedrock

AWS Bedrock gives access to every major model family through a single API. When a better model ships, I switch a config value. No new SDK, no new billing relationship, no migration. Always best-in-class, always pay-per-token.

This keeps the entire platform model-agnostic. Nothing in the codebase is tied to a specific AI provider. The context files, the chat system, the agent infrastructure — all of it works the same regardless of which model is behind it.

What's here

• Token Stack — the base platform. Auth, database, deploy, AI context files. $0.42/month.
• Token Drive — personal file storage. Photos, documents, archives. S3-backed, zero idle cost.
• Token Chat — AI conversations with streaming, history, model switching.
• Token Forums — threaded discussions stored as JSON. Human-readable and machine-readable.
• Token Agent — account provisioning, context administration, MCP endpoint for desktop AI tools.

How I build

Every project ships with a tokenburner.md context file — the full spec of the project in a format any AI coding assistant can consume. Schema, conventions, deployment target, current state. A new session with just the repo and the context file can make correct changes from the first prompt.