What is Engineering Corpus?

A library of open-source repos you choose, and research agents that read it for your coding agent. Before you design something, an agent finds how independent projects solved the same problem and reports the lines at a pinned commit with the license of every file. It works with Claude Code, Codex, Cursor and other agents, and the engine is MIT and runs on your machine.

What is a prior-art brief?

One page from one agent run showing the approaches real projects take to your problem, the one that fits your constraints, real call sites of it, and the license of each source. It ends with a receipt for every source.

What leaves my machine?

A search or a read sends your query to nobody: both run on your machine. The engine contacts the network only when indexing repos from GitHub, downloading the embedding model, or running the benchmark command.

Not yet. The engine will be open source under MIT from its first release, v0.1, on 2 December 2026, and will run on your machine. The hosted service, with the full corpus re-indexed as the projects change, is the paid part. The waitlist is for the beta invite.

Which platforms does it run on?

The release builds wheels for Linux on x86_64 and aarch64 (glibc 2.28 or newer), macOS on Apple silicon, and Windows on x64. It needs Python 3.11 or newer.

AI Coding Assistants

Engineering Corpus

Coding agents that cite open-source prior art

0 votes. Log in to give Engineering Corpus a nod Follow the workshop. Log in to follow Engineering Corpus's workshop

Community votes: 0
DR by Ahrefs: 4 no change since last check

Ahrefs Domain Rating benchmarked 13 Nov 2026, rebuilt 13 Nov 2026.

A library of open-source repos with research agents that find how projects solved your problem, cite the exact lines at a pinned commit, and report the license

What it is

Screenshot of the Engineering Corpus homepage — engineeringcorpus.com, as visitors see it

Engineering Corpus gives coding agents access to a curated library of open-source repositories you select, along with research agents that examine the library before building. These agents identify how established projects solved the same problem, cite the exact lines at a specific commit, and report the license class of every file.

Before designing a solution, an agent runs a prior-art search to find relevant code. The result is a one-page brief showing the approaches real projects take, which one fits your constraints, actual call sites, and the license of each source. Every report ends with a receipt listing the repo, commit, lines, license class, and whether you took the idea or copied the code.

The engine is MIT licensed and runs on your machine. It works with Claude Code, Codex, Cursor and other agents as an MCP client. The hosted service, with the corpus re-indexed as projects change, is the paid part.

What it is built with

Cites open-source code at specific commits with exact line numbers
Prior-art briefs showing how projects solved the same problem
Automated license classification and verdicts
Semantic search using embedding model and BM25
Works with Claude Code, Codex, Cursor and other MCP clients
Runs locally on your machine under MIT license
Receipt with repo, commit, lines and license class

Pricing

Pricing model: Freemium

Open source engine (MIT); hosted service is paid

Follow the workshop. Log in to follow Engineering Corpus's workshop

Following Engineering Corpus means an email when they reach a milestone, post an update or launch a new product.