CodeCora: Building a Software and AI Studio in the Open

CodeCora: Building a Software and AI Studio in the Open

We build developer tools for AI agents. Everything is open source. This is who we are and how we work.

What we build

CodeCora is a software and AI studio building tools that make developers faster. We focus on the intersection of AI and developer experience: memory for AI agents, code intelligence, vectorization, and developer infrastructure.

Our products:

Product | What it does | Stars | License

Uteke | Local-first semantic memory engine for AI agents | 150 | Apache 2.0

Cora Code | CLI-first AI code review with Brain Mode | 14 | MIT

Covecto | Dual-engine image-to-SVG vectorization | 2 | MIT

DrawOver | macOS screen annotation tool (Rust + Tauri 2) | 1 | AGPL + CLA

TrapFall | Self-hosted error capture engine (Sentry-compatible) | 4 | MIT

Hompimpah | Educational game platform (SvelteKit + Phaser) | 0 | MIT

Most of our tools are written in Rust. We chose Rust for performance, safety, and distribution. When your tool runs locally on a developer's machine, binary size and startup time matter. Rust delivers on both.

Why open source

Every product we ship is open source. The code lives at github.com/codecoradev. No source-available, no fair-code, no delayed releases. The source is the release.

The reasons are straightforward. Developer tools should be transparent. If you run a code reviewer on your codebase, you should be able to read every line of it. If you store memories in an engine, you should know exactly how it works. Closed source developer tools ask for blind trust.

We also build for developers who want to own their infrastructure. Uteke runs on your machine. Cora Code runs in your terminal. Covecto runs on your server. No cloud dependency, no API keys to our servers, no vendor lock-in.

Then there is distribution. Open source is the most efficient distribution model for developer tools. A developer who finds your repo, reads the README, and runs cargo install is worth more than 100 signups on a landing page.

License strategy

We use three different licenses based on product type:

  • Apache 2.0 for infrastructure (Uteke). The patent grant matters when your tool sits in the middle of someone's stack. Apache 2.0 gives enterprises confidence to adopt.
  • MIT for developer tools (Cora Code, Covecto, TrapFall). Simple, permissive, no friction. Developers should be able to use your tool without reading a license.
  • AGPL + CLA for end-user apps (DrawOver). When the product is a complete application, AGPL prevents cloud providers from wrapping it as a service without contributing back.

This is not a philosophical stance. It is a practical decision per product.

How we work

We are a small team. Bootstrapped, no VC funding. This shapes how we build.

We ship fast. Uteke shipped 40+ releases in 58 days. We ship small, ship often, and fix fast. No quarterly release cycles.

Every PR runs through checks: format, clippy, test, build. If CI fails twice, we stop and fix before merging. No "fix it later" merges.

Documentation is a feature. Covecto shipped with 14 pages of VitePress docs on day one. Uteke has API docs, getting started guides, and integration examples. If a feature is not documented, it does not exist.

Every PR is reviewed and merged by a human. No auto-merge. This is slower but catches issues that CI cannot.

What is next

We are in early days. Uteke is the most adopted product (150 stars, 40+ releases). Cora Code and Covecto are newer. DrawOver and TrapFall are even earlier.

The plan for the rest of 2026:

  • Uteke v0.11+: better multi-agent coordination, improved recall performance, more language SDKs
  • Cora Code v0.9+: more deterministic rules, improved Brain Mode accuracy, IDE integrations
  • Covecto v0.2+: more vectorization profiles, improved Spline engine, batch API improvements
  • DrawOver: public beta, Polar.sh integration for freemium model
  • TrapFall: first stable release, Sentry SDK compatibility

We build in the open. You can watch every commit, every release, every issue. If you want to try our tools, start with Uteke (brew install codecoradev/tap/uteke) or Cora Code (cargo install cora-code).

Star a repo, open an issue, send a PR, or just read the code.

That is how we build.

*CodeCora is a software and AI studio under PT Azfirazka Digital Kreatif. All products are open source at github.com/codecoradev. Docs at codecora.dev.*