Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

JCR (Jev Capability Resolver)

Gives an agent one tool that uses Jev to search a nested capability tree and return only the documented commands a task needs.

Projectprototype
View on GitHub ↗Live demo ↗

Overview

JCR (Jev Capability Resolver) moves command lookup out of the agent's context. Instead of an agent reading several skill or documentation files to find a few commands and then carrying them through the task, it calls one tool; JCR uses Jev to walk a nested capability tree and returns the context attached to the selected operations. It does not execute commands.

The author proposes a provider-independent capabilities format for deterministic commands, complementing skills that describe workflows and judgment, and would like it to become an open standard. The repository includes the resolver, a capability catalog, a stdio MCP server, comparison harnesses for Claude and Codex in skills mode versus JCR mode, and 50 benchmark scenarios. An OpenAI key handles compound requests, and an article and demo are on the project site.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich branch of the capability tree is relevant to this task?(inferred)—
Yes / noIs this operation needed for the task?(inferred)yes · no

How it works

  1. Agent calls the JCR tool with a task description
  2. Jev selects branches and operations in the capability tree
  3. Compound requests are split with an LLM
  4. JCR returns only the documentation for selected operations

Strengths

  • Keeps search out of the main agent's context
  • Works over MCP with Claude and Codex
  • Includes 50 benchmark scenarios for comparison

Caveats

  • Needs TypeSafe and often OpenAI keys
  • Capability format is a proposal, not an established standard

Getting started

Node 22+: `git clone https://github.com/NiazMorshed2007/jcr.git && cd jcr && npm ci && cp .env.example .env`, then fill in keys.

At a glance

Type
Project
Status
prototype
Language
TypeScript
Runtime
Node, stdio MCP server
Has code
yes

Use cases and integrations

How curated lists describe it

  • One tool that searches a nested capability tree and hands the agent only the documented commands and context a task needs.

    MrJev/awesome-jev

Frequently asked questions

What does JCR decide?

Which capabilities and operations in a nested tree are relevant to the agent's task.

Does it run commands?

No. It only returns documentation for the selected operations.

Which agents can use it?

Any MCP client; harnesses compare Claude and Codex.

Related projects

  • JevScout

    Toolkit where Jev condenses large MCP responses for Claude Code and maps Codex requests to reviewed, fixed repository operations.

    ProjectbetaTypeScript
  • jev-skill-router

    Local Python MCP service that keeps the skill catalog out of the agent's context and lets Jev pick which SKILL.md files to load.

    MCP serverbetaPython
  • pi-jev-skill-suggestion

    Pi extension that strips the inlined skill list and asks Jev which single skill, if any, to load per turn.

    Agent skillbetaTypeScript
  • jev-awesome-skills

    Open-source Jev skills for Claude Code, Codex, Cursor and Grok: one client to gate, route, guard, verify, triage, review and act.

    ProjectbetaTypeScript
  • jev-classifier

    Local gateway and MCP tools that let Jev predict which tool a coding agent should call next, from suggestions to enforced choices.

    MCP serverbetaTypeScript
  • jev-skill-selection

    Pre-message hook that keeps or drops SKILL.md files before the first model call, using local overlap or Jev keep/drop judgments.

    HookbetaPython