User repository on GitHub
Claude Code mod: per-prompt reasoning effort via Jev (OpenRouter)
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Integrations & Adapters
Jev Effort
A Claude Code mod that sends each prompt to Jev through OpenRouter and turns the returned 0-3 score into a reasoning effort level, from low to xhigh. Shadow mode is on by default, so it only logs what it would set until you turn it off.
- Stars
- 3
- Language
- TypeScript
- Reading time
- 3 min
- Published by
- Community
- Author
- gebeer
- Last verified
- Oct 2026
- Languages
- TypeScript
- Platforms
- CLIOpenRouter
- Capabilities
- Open sourceRequires API key
Highlights
- The prompt goes to Jev through OpenRouter; a 0-3 score maps to low, medium, high or xhigh effort, and the effort holds for the turn.
- Shadow mode is on by default: the mod logs what it would set, for example 'shadow · jev 0.42 → low (conf 0.71, 212ms)', without changing the effort.
- Requires Claude Code 2.1.259 or later and an OpenRouter API key; with no key, the mod does nothing.
- Defaults: upgrade needs 0.3 confidence, downgrade 0.6, the Jev wait times out at 800 ms and it uses the model ~typesafe/jev-latest.
- Privacy: up to 2,000 chars of your prompt, 1,500 of the last reply and 3 earlier prompts; tool inputs, file contents and the system prompt are never sent.
Quickstart
bash
git clone https://github.com/gebeer/jev-effort.git
CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1 claude --plugin-dir ./jev-effortWatch out
README reviewed 2026-10-17; the README says MIT while GitHub metadata reports NOASSERTION, the only testing note is 'Tested on Claude Code 2.1.289, 2026-10-04', and no independent benchmarks exist. An effort change creates a new cache on older models.
Questions about this entry
Short answers drawn from the linked project or page.
It is a Claude Code mod that sets the reasoning effort for each prompt. It sends the prompt to Jev through OpenRouter, gets a score from 0 to 3 and maps it to low, medium, high or xhigh effort for the turn.
Score 0 is low, 1 is medium, 2 is high and 3 is xhigh. The effort stays the same for all tool calls in the turn, the mod does not touch subagents, and it does not set max; with max, a numeric budget or a model without effort, it does nothing.
Clone the repository and start Claude Code with CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1 claude --plugin-dir ./jev-effort. You need Claude Code 2.1.259 or later and an OpenRouter key, set either as OPENROUTER_API_KEY or in the plugin's settings file.
It is the default. The mod only writes what it would set into the transcript and status line; it does not change the effort. The README suggests comparing the decisions with your own opinion first, then switching shadow to false in /config.
apiKey, shadow, sendContext, minUpgradeConfidence (0.3), minDowngradeConfidence (0.6), timeoutMs (800), logDecisions, model (~typesafe/jev-latest) and endpoint. For TypeSafe direct, use https://api.typesafe.ai/v1/systemone with the model jev-latest.
A timeout, an HTTP error or a bad answer makes the mod keep the effort of the previous turn. It also skips prompts it does not classify: slash commands without text, prompts typed while a turn runs, empty prompts and messages that are not from you.
OpenRouter sends Jev up to 2,000 characters of your prompt, 1,500 of Claude's last reply, 3 earlier prompts of 300 characters each, and the last turn's tool names with counts and failure counts. Tool inputs and outputs, file contents, CLAUDE.md and images are not sent.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
A host-agnostic routing layer that sits between a coding harness and its models, choosing capacity per round from trajectory evidence like tool calls, failures and context pressure, running as a user-scoped daemon and local OpenAI-compatible proxy.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
vizuh/sabiTypeScript
vizuh/sabiTypeScriptA Claude Code mod that asks Jev one Noul question per skill for every prompt, ranks the skills above the threshold and passes them to the model as a suggestion. It never loads a skill itself, and a timeout, outage or missing key only skips the routing.
Jev Integrations & Adapters#community#github#claude-code
Community
gecm0/skill-router-modTypeScript
gecm0/skill-router-modTypeScriptA Claude Code plugin that asks Jev, TypeSafe's System One model, which model tier each prompt needs — haiku, sonnet or opus — and switches automatically, and routes classification, scoring and yes/no judgments to Jev via scripts/jev.py.
Jev Integrations & Adapters#community#github#claude-code
Community
ricky94511/jevTypeScript
ricky94511/jevTypeScriptClaude Code plugin that asks Jev a few typed questions before every turn and sets reasoning effort, subagent model and effort, strategy, and the skill a prompt needs; you keep Opus for the conversation.
Jev Agents & MCP#community#catalog-import#logicrw
Community
akramovic1/jev-pilotTypeScript
akramovic1/jev-pilotTypeScriptA Claude Code plugin that sends every Agent spawn through JEV (TypeSafe), which picks the model (haiku, sonnet, opus), the effort and the tool profile of each subagent. It fails open to sonnet at medium effort, accepts a TypeSafe or OpenRouter key, and ships plan and run skills.
Jev Agents & MCP#community#github#claude-code
Community
hiyoku/jev-orcTypeScript
hiyoku/jev-orcTypeScriptAn OpenCode V2 server plugin that uses Jev to route requests between the Plan, Build and your own primary agents, and hand off unfinished work at completed execution boundaries while following each target agent's configured model.
Jev Integrations & Adapters#community#github#typescript
From the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.