stanley-code
Jev-first CLI coding agent that routes a request to a deterministic workflow which asks Jev fixed-choice questions about bounded evidence.
Overview
Stanley (formerly the jev-code placeholder) is a command-line tool for checking and understanding code changes. You describe what you need, and Jev routes the request to one workflow: built-in, a trusted repository workflow, or one Stanley wrote for itself. Each workflow is deterministic code that gathers bounded evidence and asks Jev small fixed-choice questions; code applies fixed thresholds and makes the decisions, and this hot path never starts a general agent. Routing requires confidence of at least 0.6, probability at least 0.55, a margin of 0.15 and an available workflow, and includes a cannot_tell option.
Built-in workflows cover review, test gaps, summaries, security, performance and compatibility. For example, when checking whether changes fix a crash, Jev asks whether each changed hunk relates to the request and whether a test expectation weakened, and it relates CI failures to the diff as caused, unrelated or environmental. Reports separate findings from what was not checked, and an empty findings list is not an approval. When no workflow fits, it falls back to the Pi coding agent and then has an agent draft a new workflow for you to promote. Version 0.1.0 is not yet on npm.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which workflow should handle this request (or cannot_tell)? | — |
| Yes / no | Is this changed hunk related to the request? | — |
| Yes / no | Did a test expectation become weaker? | — |
| Choice | Is this failure caused by the diff, unrelated, or environment/infrastructure? | caused by diff · unrelated · environment/infrastructure |
How it works
- Describe the request in plain language
- Jev routes it to an eligible workflow if confidence gates pass
- The workflow gathers bounded evidence and asks Jev fixed-choice questions
- Code applies thresholds and reports findings and not-checked items
Strengths
- No general agent on the hot path
- Explicit routing gates and cannot_tell option
- Self-extending via staged workflows
Caveats
- Not yet released on npm
- Empty findings are not an approval
- Windows untested
Getting started
git clone https://github.com/devagrawal09/stanley-code.git && npm ci && npm run build, export TYPESAFE_API_KEY, then node dist/cli.js "your request".At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- Node CLI
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Coding agents: bounded Jev workflows that keep agent judgments typed instead of free-form.
yibie/awesome-jevBounded TypeSafe Jev workflows for coding agents.
hellogumbo/awesome-jevBounded Jev workflows for coding agents (formerly jev-code).
MrJev/awesome-jev
Frequently asked questions
What does Jev decide in Stanley?
Which workflow handles a request, and fixed-choice questions about each piece of evidence.
Who makes the final call?
Code, using fixed thresholds.
What happens if no workflow fits?
It falls back to the Pi agent and drafts a new workflow for you to promote.
Related projects
Software-factory supervisor where Jev independently judges coding agents' completion, test sufficiency, progress and need for a human.
ProjectbetaPythonRust code-review gate asking Jev small typed questions about functions and tests, failing CI only on rules measured 80% right.
CLIproductionRustLocal desktop pal that verifies what a coding agent really did, with optional Laya or Jev double-checks of unclear test results.
ProjectbetaTypeScriptClaude Code plugin with rules and checks that make agents prove their work; an optional Jev judge triages review findings.
PluginbetaMarkdown + shellOpenAPI diff tool that adds Jev semantic review of changed prose to deterministic checks, then code decides pass, review or block.
ServicebetaPythonClaude Code plugin that scores a turn's redacted diff on eight Jev risk axes at Stop and asks Claude to reinvestigate once if high.
ProjectbetaGo