What this is

We Can Just Build Things is an open operational system for building production-grade, privacy-respecting software with AI agents — without having to already be a developer to make the decisions that matter.

It exists because people with ideas hit a wall the moment they try to act on them. Not a wall of capability — a wall of navigation. There is no shortage of free, open-source tools. There are tens of thousands of them, scattered across registries and Discord screenshots, with no reliable way to tell which are maintained, which quietly relicensed last year, and which route your users’ data through infrastructure you’d never knowingly choose.

This project solves both problems at once: it makes the landscape navigable for someone who isn’t already inside it, and it takes a position on what belongs in it — enforced by tooling, not by good intentions in a README.

The three things it gives you

A verified catalog Tools license-checked at a specific commit and screened against a strict exclusion policy. Each entry shows its work.

A guided build flow A decision flow adapted from the PIE PDX accelerator: from “my community has a problem” to a concrete, vetted plan.

A policy engine with teeth A three-layer check that blocks anything depending on Meta, OpenAI, or xAI — directly or transitively.

Reusable agent skills Skills and recipes that encode real engineering discipline into repeatable workflows for Goose or Claude Code.

How you’ll actually use it

1. Start from your problem, not a tool. Walk the build flow to turn an intent (“my community needs to document evictions safely”) into a shape: what you have, what’s missing, what the working theory is.

2. Get a vetted shortlist. By the end you have concrete, license-checked components that match your stage, your stack, and the values you care about.

3. Hand the frame to an agent. Point Goose, Claude Code, or Aider at the recipes and skills. The agent runs inside a set of decisions that have already been made, verified, and documented.

4. Ship something accountable. The output is a stable artifact someone can audit — pinned dependencies, a clean provider posture, a real license.

You don't have to become a developer

The person defining what their community needs never has to become an engineer. But the product they deliver should be indistinguishable from one built by a professional team. That gap is what the catalog, the flow, and the policy engine close.

Who it’s for

This is not built for the founder who already controls the agent network and sets the norms. It’s built for the organizer, the nonprofit, the feminist tech collective — the person who has to trust a system without being able to see inside it. Read why that distinction matters →