jevbooks

← All projects

reflex-computer-use

8itlab/reflex-computer-use · Homepage

Computer-use MCP server that reads the UI tree and uses Jev as a fast judge.

Agent Decisions0.96Runner-up: Evaluation & Benchmarking
Automated README summary

Automatically summarized from the README on 3 Nov 2026; not a run by jevbooks. Run instructions and author-reported outcomes still need verification.

What it is

An open-source computer-use MCP server for Claude Code, Codex and other AI agents. It controls local macOS, local Windows, or remote Windows over SSH by reading the accessibility/UI Automation tree instead of screenshotting every step.

How it uses Jev

Jev powers find (match an element from a one-line description), check (yes/no question about the screen), and wait (poll until true or timeout). It judges from the read UI text by default; with look:true it also sees the window screenshot to locate icons and game objects. Results return a match and confidence used to click or gate steps.

Primitives:choicescore

Technique worth stealing

Read the accessibility/UI tree for element coordinates instead of screenshotting each step, and reserve the fast model for small high-frequency judgments.

Try it

Register with claude mcp add reflexcu -s user -- uvx --from <GitHub tarball URL> reflex-computer-use, then set a Jev key.

View on GitHub

Start using Jev: API key and your first request →

Model judgments from README excerpts

Project signals use repository metadata, an existing summary and the first 700 README characters, not the full project. “Not found” means not detected in that material; it does not establish absence. These are the original judgments with recorded models and dates, not source audits or run results.

Models & evaluation dates

Inclusion gate
jev-1.14.0
Project signals
jev-1.14.0
Category
jev-1.14.0
Pattern tags
jev-1.14.0

Evidence

These signals use repository metadata, an existing summary and at most 700 cleaned README characters. ≥ 0.60 means reported in that excerpt; ≤ 0.40 means not found there. The middle band is inconclusive. None is a source-code audit or a test result. The runnable-tool question excludes research experiments; the recommendation question asks for a learning shortlist, not a quality verdict.

  • Jev is core to the projectnot found in the excerpt0.35
  • Shows a System One patternnot found in the excerpt0.33
  • Describes handling uncertaintynot found in the excerpt0.03
  • Reports measured numbersnot found in the excerpt0.03
  • Is a runnable toolFound in material0.97
  • Top-10 learning shortlist signalnot found in the excerpt0.30
  • Is a replica of the modelnot found in the excerpt0.03
  • Solves a common problem (0–2)score on a 0–2 scale1.06 / 2
  • About JevFound in material0.99
Signals
Jev is core to the project0.35Shows a System One pattern0.33Describes handling uncertainty0.03Reports measured numbers0.03Is a runnable tool0.97Top-10 learning shortlist signal0.30Is a replica of the model0.0300.51Solves a common problem (0–2)1.06 / 2012

no-judgment band (0.40–0.60)

Design patterns
Fail open, fail closed0.67Funnel0.65Blind review0.47Check-up0.32Menu0.28Inspection0.2800.51

no-judgment band (0.40–0.60)

The Jev model and date for each judgment stage are listed above. Card written by DeepSeek V4.1 Flash from the README on 3 Nov 2026.