User repository on GitHub
Bounded TypeSafe Jev workflows for coding agents.
- Forks
- 6
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Agents & MCP
Jev Code — devagrawal09
Stanley is a CLI where you describe a code-review request in plain language and Jev routes it to one of ten bounded workflows. Each workflow is deterministic code that gathers bounded evidence and asks Jev fixed-choice questions per hunk; unmatched requests go to the Pi agent.
- Stars
- 116
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- devagrawal09
- Last verified
- Oct 2026
About this project
Stanley is a command-line tool for checking and understanding code changes. Instead of naming a command, you describe what you need. TypeSafe Jev routes the request to one workflow: a 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 about it; code, not a model, makes the decisions. That is the hot path, and it never starts a general agent.
Quoted from the project's README. Read the full README.
- Languages
- TypeScript
Highlights
- Jev only answers fixed-choice questions about one small piece at a time; thresholds and evidence choice live in code.
- Routing reads author-provided JSON metadata plus deterministic facts like whether a diff exists or input is a failure log.
- Picks below confidence 0.6, probability 0.55 or margin 0.15 end in cannot_tell instead of a guess.
- Exact signals are recorded without asking the model: added test.skip, deleted assertions, deleted test files, lockfile or CI edits.
Watch out
stanley-code is not on npm yet — the README describes an unreleased 0.1.0, so it has to be built from source, and the older jev-code package on npm is 0.0.1 and does nothing. Windows is untested and CI runs on Linux. The API key is read only from TYPESAFE_API_KEY in the environment, never from files or flags. Work is bounded by policy: the task check judges the first 150 hunks, and find screens the top 3000 files by lexical score and returns ten. The README labels the whole thing experimental and says any request, report or interface may change. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
Not usefully. stanley-code is not published yet — the README describes version 0.1.0 as unreleased and tells you to clone, run `npm ci` and `npm run build`, then invoke `node dist/cli.js`. After 0.1.0 ships it would be `npm install --global stanley-code`. The old jev-code placeholder package that shares the name on npm is version 0.0.1 and does nothing.
No, and the README says so directly. Output carries both `output.data.findings` and `output.data.notChecked`, and an empty findings list is explicitly not an approval. Bounds also apply: the task check judges the first 150 hunks by policy, generated and unreadable files are excluded, and hunk priority is set by deterministic caps rather than by the model.
Only classification. Built-in code decides which hunks matter, applies fixed thresholds to the probabilities Jev returns and combines them; Jev answers multiple-choice questions about one small piece at a time. Routing needs confidence at or above 0.6, probability at or above 0.55 and a margin at or above 0.15, and a pick that misses those returns cannot_tell rather than running a workflow.
Two fallbacks. If the Pi coding agent is installed, Stanley delegates to it, reports the agent's own account of what it did without claiming to have verified it, and in the background asks an agent to draft a Stanley workflow for that kind of request, which you validate, stage and promote. With no agent installed you get a clarification carrying status unsupported. Trusted repository workflows under .stanley/workflows/ can also cover the gap directly.
Find relevant code, check current changes, review current changes, find test gaps, summarize current changes, review security, review performance, review compatibility, triage test failures, and triage review comments, with repository workflows under .stanley/workflows/ extending them.
Exact signals from the diff: an added test.skip, deleted assertions, deleted test files, whitespace-only edits, and lockfile, CI, or config changes are recorded as observations rather than put to Jev.
When no workflow supports a request it falls back to the Pi agent if installed, and in the background asks an agent to write a new Stanley workflow; the candidate is validated and staged, and you promote it so the next such request runs with no agent.
Other listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
45GitHub stars
Five typed tools — classify, check, score, rank, and ask — plus a skill, wired into Claude Code, Codex, Pi, and OpenCode in one setup command.
Jev Agents & MCP#community#typescript#mcp
Community
FrancoisChastel/jev-codeTypeScript
FrancoisChastel/jev-codeTypeScript20GitHub stars
TypeScript SDK and CLI coding agent where Jev only picks options while your code runs effects, building programs as syntax trees.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
rhighs/jev-codeTypeScript
rhighs/jev-codeTypeScriptMore Jev resources like this
1.5kGitHub stars
Open-source platform that puts ACP-compatible agents such as Claude Code, Codex, and Pi inside the team's Slack, Telegram, Discord, and GitHub workflows, with reusable Jev decisions for routing, response gating, and reviewer selection.
Jev Agents & MCP#community#jevbest
488GitHub stars
A coding agent built on pi that hands its non-code decisions — what enters the context, whether a command is safe, whether the work is done — to a small judge model at 35 decision points per turn.
Jev Agents & MCP#community#jevbest
Community
qybaihe/muTypeScript
qybaihe/muTypeScript210GitHub stars
An agent plugin for Pi, Claude Code, Codex CLI and Grok Build that asks Jev two classification questions to decide whether a session sits at a safe compaction boundary, then hints or compacts for you.
Jev Agents & MCP#community#jevbest
572GitHub stars
A CLI plus agent hooks that ask Jev one question per rule in your AGENTS.md on every edit or turn and return a probability, so a violating edit gets repaired inside the same turn. Claude Code, Codex, OpenCode, Pi.
Jev Agents & MCP#community#jevbest
Community
coldteadotai/abideTypeScript
coldteadotai/abideTypeScript147GitHub stars
A local-first desktop agent for macOS that carries work from a conversation through to a reviewable result, split into Home, Cowork and Crew. The part worth a developer's attention is Crew, where Jev suggests an assignee for a single unassigned task and you accept or ignore it.
Jev Agents & MCP#community#jevbest
Community
Foxtailsss-Andy/Anna-AgentTypeScript
Foxtailsss-Andy/Anna-AgentTypeScript48GitHub stars
Agent harness for work that outlives one chat, with resumable turns, routed state machines, and observational memory across sessions and processes.
Jev Agents & MCP#community#jevbest#catalog-import
Community
dzhng/duet-agentTypeScript
dzhng/duet-agentTypeScriptFrom the Jev community
Posts from builders shipping with Jev right now.