What it is
jeffrey is a coding-agent CLI for developers. A decision model picks the next tool and scores progress, risk, and goal completion; an OpenAI-compatible LLM fills in the tool arguments and does the actual work.
How it uses Jev
On every step Jev answers closed questions: which tool to call, how much progress was made, estimated risk, whether the goal is reached, and a stuck probability. The agent uses these to pick the next tool, gate mutating tools on risk >= 0.5, and escalate when stuck.
Primitives:choicescorenoul
Technique worth stealing
Split deciding from executing: a closed-question decision model chooses tools while a separate LLM only fills arguments.
Try it
npm install -g @thomasbrueggemann/jeffrey && jeffrey "add retry with backoff to the http client"
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricno0.37
- Shows a System One patternyes0.81
- Handles uncertaintyno0.06
- Measuredno0.05
- Runnableyes0.93
- Worth recommendingno0.35
- Model replicano0.07
- Problem scopescore on a 0–2 scale1.42
- About Jevyes0.96
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.