User repository on GitHub
Local Paseo plugin exposing Jev evaluations through MCP
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Integrations & Adapters
Hiep Paseo Plugin
Fifteen independent local plugins for the Paseo desktop app, two of which use Jev: an evaluator exposed to Codex and Claude agents over MCP, and an orchestrator that routes a task to one model profile and effort level.
- Stars
- 3
- Language
- TypeScript
- Published by
- Community
- Author
- HiepPP
- Last verified
- Oct 2026
About this project
See what your agents are doing. Keep projects organized. Bring task context and workspace checks into Paseo without leaving your conversation. Fifteen independent local plugins. Install only what you need; each owns its dependencies, checks, and configuration.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- Each plugin owns its dependencies, checks and configuration; there is no root workspace build to run.
- Board groups conversations into pastel project blocks with bot avatars, parent/subagent clusters and 10% to 200% UI scaling.
- Workspace Spaces adds numbered sidebar tabs; removing a Space moves its projects to a neighbour and the last one is protected.
- Workspace preflight reports unknown coverage separately from passes, so missing configuration never looks like a green check.
Watch out
The gallery images are AI-edited screenshot illustrations with conversation titles, project names, profile notes and local paths replaced by sample text, described as visual previews rather than pixel-exact test evidence, and the Next prompt actions image comes from an earlier desktop build. Jev reaches Paseo through MCP, not a standalone page, and the evaluator capture shows installation status rather than an evaluation result; evaluations need a configured Vercel AI Gateway key and consume API quota. Next prompt actions auto-run is optional, starts off, is desktop only and uses a private adapter for the existing Markdown blocks, while Spaces has a standalone fallback elsewhere. Thread janitor archives idle threads past 24 hours with no confirmation dialog, though the sweep never deletes agents, files or worktrees. The README states that installed or running status alone does not prove every plugin action works, and loop-verify is labelled a prototype capped at 5 rounds. 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.
The README says the opposite. The refreshed images are AI-edited illustrations that crop the UI and swap private titles, project names, profile notes and local paths for sample text, and it calls them visual previews rather than pixel-exact test evidence. Several captures deliberately show an unconfigured state, such as an empty task field or Light still selected in the theme picker.
API quota, through the Vercel AI Gateway key you configure. The evaluator exposes Jev as a typed evaluation tool with boolean, choice and score questions over state you choose to share, so you decide what goes into the request. Workspace preflight pairs with it to evaluate which evidence matters, and your coding agent stays in control rather than having the evaluation drive the edit.
Yes, that is the stated design. You cd into the plugin directory, run npm ci, npm run typecheck, npm run lint and npm test, then paseo plugin install "$PWD" --id <plugin-id>. Each plugin declares its own paseo-plugin.json and runtime entry, no root workspace build is required, and after editing one you reload it with paseo plugin reload.
Not according to its documented behaviour. By default a thread idle for more than 24 hours is archived, archiving stays reversible, and it never deletes agents, files or worktrees. Sweeps run after agent.turn_ended and agent.created hooks plus a 30-minute timer, at most once per 10 minutes. The catch is that there is no confirmation dialog.
vscode-warm-light, claude-look, next-prompt-actions, prompt-translate, skill-pins, thread-branch, jev-orchestrator, Board, jev-evaluator, watchtower-board, workspace-preflight, workspace-spaces, thread-janitor, thread-context-attach, and loop-verify, which is labelled a prototype capped at 5 rounds. Each is a self-contained directory with its own paseo-plugin.json and runtime entry, and no root workspace build is required.
Several, plainly. Next prompt actions auto-run is optional, starts off, is desktop only, and uses a private adapter. Spaces has a standalone fallback on other clients. Thread janitor archives with no confirmation dialog. PR discovery in thread-branch needs an authenticated gh while the branch itself only needs Git. And the gallery images are AI-edited illustrations, with the next-prompt-actions capture coming from desktop 0.8.0 while the rest are from 0.9.0-beta.2.
Its Direct Jev routing panel lets you pick saved profiles and allowed effort levels before Route and run creates one agent in the current workspace, with profile permissions and features staying attached to that choice. It also offers optional scoped delegation with discovery, escalation, review, and ownership locks. Open it from Command Center or /route, and routing consumes API quota.
More Jev resources like this
69GitHub stars
Provider-neutral TypeScript runtime for System One decisions: write choice, noul, and score code once, then run it on TypeSafe Jev, local Reflex, any compatible /v1/systemone endpoint, or a custom provider.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
iamaamir/system-oneTypeScript
iamaamir/system-oneTypeScript21GitHub stars
Language and VM for agent programs that wait for approval, survive restarts, and replay portable receipts without recalling the model.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
hraness/algalTypeScript
hraness/algalTypeScript3GitHub stars
TypeScript guardrail that puts seven Jev checks in front of an AI SDK model or an OpenAI-compatible endpoint, blocks on high-confidence hard checks, downgrades low confidence to a flag, and logs every verdict as a PII-free JSONL row you can turn into an eval set.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
noelzappy/tripwireTypeScript
noelzappy/tripwireTypeScript431GitHub stars
Smithers turns issues into reviewed, tested changes through flows defined beside your code, with an optional step that keeps the project wiki current. The 1.0 release candidate runs from a source checkout.
Jev Integrations & Adapters#community#jevbest
Community
smithersai/smithersTypeScript
smithersai/smithersTypeScript38GitHub stars
File-first project notebook where many agents share context in .context/ files while Jev only judges the next step and records it as JSON.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
pedrocivita/tocketTypeScript
pedrocivita/tocketTypeScript34GitHub stars
Self-hosted agent runtime that adds an optional Jev judgment layer to memory filtering, lesson dismissal, request classification, and skill selection, with inspectable context and execution.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
matrixorigin/AstraRust
matrixorigin/AstraRustFrom the Jev community
Posts from builders shipping with Jev right now.