jcr

# JCR · Jev Capability Resolver

[](https://jcr.niazmorshed.dev)

JCR gives an agent one tool to find the documented commands it needs for a task.
It uses [Jev](https://typesafe.ai/blog/introducing-system-one-models-and-jev) to
search a nested capability tree and return the context attached to selected
operations.

Skills made it easier to load instructions when they are needed. But an agent
can still read several files to find a few commands, then carry those files
through the rest of the task. JCR moves that lookup into a resolver. The main
agent gets the selected instructions while the search stays outside its context.

I built this around a proposed capabilities format for **deterministic commands**
from any provider. Skills can describe a workflow and the judgment it needs.
Capabilities can document the individual operations used along the way. The
format is independent of Jev, and I would like to explore whether it should
become an open standard with the community.

**[Read the article and watch the demo](https://jcr.niazmorshed.dev)** ·
[Quick start](#quick-start) · [Capability format](#capability-format) ·
[Benchmarks](#benchmarks) · [Contributing](#contributing)

This repository includes the resolver, a capability catalog, a stdio MCP
server, Claude and Codex comparison harnesses, and 50 benchmark scenarios.
JCR returns documentation. It does not execute commands. The included
harnesses also stop at explaining the steps needed to carry out a task.

## Quick start

Use Node.js 22 or newer and npm. Clone the repository and install its dependencies.

```bash
git clone https://github.com/NiazMorshed2007/jcr.git
cd jcr
npm ci
cp .env.example .env
```

Run the remaining commands from the repository root.

Fill in the keys in `.env` for the mode you want to run. For a