Open source · Rust · MIT
I publish the libraries and tools I build for my own apps, agents and games. These six are the ones most likely to save you a week. Every README example compiles, every crate states its minimum Rust version, and issues get answered.
69crates on crates.io5,038downloads3,900in the last 90 days
Server, CLI and library · v2.0.2 · MIT
Stop paying twice for the same LLM call.
A proxy in front of any LLM API. The same prompt asked twice costs one model call, outages fail fast with a clear reason, and a spend cap stops runaway bills. Point an OpenAI or Anthropic client at it and change nothing else.
cargo install tokio-prompt-orchestrator --features web-api159 downloads · 88 in the last 90 days
Library · v2.15.1 · MIT
Trading code that refuses bad numbers.
Exact decimal prices and sizes, a level-2 order book, ticks to candles, 700+ streaming indicators with TA-Lib definitions, Black-Scholes Greeks, a position ledger and risk limits. Works next to ta, yata and wickra.
cargo add fin-primitives573 downloads · 270 in the last 90 days
CLI, terminal UI and web UI · v5.0.1 · MIT
See how sure the model is about every word.
Watch an LLM answer token by token with the confidence and perplexity of each token from logprobs, then rewrite tokens while it streams. OpenAI, Anthropic, Gemini, Ollama. A mock provider needs no API key.
cargo install every-other-token162 downloads · 68 in the last 90 days
Library and demos · v0.3.0 · MIT
Write the equations, watch them move.
Real-time graphics from math: strange attractors, ODEs and force fields drawn as glowing particles with HDR bloom on OpenGL. 21 runnable demos and a screensaver.
cargo add proof-engine164 downloads · 129 in the last 90 days
CLI, dashboard and library · v1.3.0 · MIT
Know what every LLM call cost.
Price each call from your own logs or the provider response: 104 models, prompt caching, Claude Code session logs. Watch it live in a terminal dashboard, scrape it with Prometheus, or call it from Rust.
cargo install llm-cost-dashboard126 downloads · 67 in the last 90 days
Library · v0.2.0 · MIT
A hard dollar cap for every agent.
Charge each model call to a per-agent budget in US dollars and get a typed error the moment a call would cross it, plus a kill switch and an automatic step down to a cheaper model. 232 real model prices built in.
cargo add llm-budget54 downloads · 31 in the last 90 days
Every crate: crates.io/users/mattbusel. Source: GitHub, mirrored on GitLab.
Open an issue on its GitHub repo, even for a question; that is how I find out what to fix first. If one of them ends up in your project, I would love to hear about it: matt@tensorust.com. Need something built on top of them? See Build.
What the download numbers do and do not mean, measured on these same 69 crates: crates.io download count: 80% of ours arrived with a new version.
Download counts from the crates.io API, fetched when this page was built (Oct 17, 2026). crates.io counts every fetch of a .crate file, including CI runs and mirrors, so it is a rough signal, not a user count.
Get the receipts by email.
What my AI agents tried this week, with the real numbers: what worked, what flopped, what made money. One email when there is something worth reading.
No spam, never sold. Every email has a one-click unsubscribe.
Browse: Bundles and kitsHome, family and schoolMoney and businessAI toolsSports and fitnessMaking and hobbiesFood and drinkIndie dev and launchTrading and quant
Free: Landing page teardown iPhone app cost calculator LLM API cost calculator Paid online: $49 Launch Check App builds Articles FAQ