# Flynn

> Flynn is a secure, self-improving agent operating system in a single Go binary. Bring your own model, point it at a goal, and grant it real authority: every action is sandboxed, governed, and sealed into a verifiable, tamper-evident record.

Flynn runs standalone as one static binary from a laptop to a $5 VPS, works with any model provider (hosted or fully local), and stores all of its state locally, so you own it. It is the reference implementation of Provetrail, the open standard for verifiable execution provenance. Apache-2.0 licensed.

## Core

- [Homepage](https://flynnhq.com/): overview, feature summary, and community signup for builders, founders, engineers, and researchers.
- [Flynn on GitHub](https://github.com/ionalpha/flynn): source, one-line install, documentation, and roadmap.
- [Flynn: the reference implementation of Provetrail](https://flynnhq.com/posts/flynn-the-reference-implementation/): deep dive on how Flynn produces a governed, tamper-evident record from inside the work, and how anyone can verify it.

## Standard

- [Provetrail](https://provetrail.org): the open standard for verifiable execution provenance that Flynn implements and passes across every conformance tier.

## Key capabilities

- Run any model, including fully local open-weight models, with hardware-fit checks and no lock-in.
- One static binary, zero runtime: no Python, no Node, runs from a laptop to a $5 VPS.
- Hard per-run token and cost ceilings, so runaway spend is structurally impossible.
- Capability-scoped, sandboxed, governed actions sealed into a verifiable record you can check.
- A closed learning loop that captures skills and memory from real work and reinforces what succeeds.
- Goals-and-missions orchestration that turns an instruction into a plan and fans it out into concurrent, governed runs.

## Optional

- [Ion Alpha](https://ionalpha.io): the team building Flynn in the open.
- [Privacy policy](https://flynnhq.com/privacy/): how flynnhq.com handles email signups and data.