Triage 2026-09-26: the top 30, all 30 listed

[MrJev]

Fourth day under the queue rule from #45. Discovery found 43; this is the top 30 by stars, and the 13 rows below the line keep their place for tomorrow.

Every candidate was shallow-cloned and grepped in code. Unusually, nothing was ignored and nothing was held back: all thirty either made a real call or were open decision models, and all thirty have a README that explains how to run them.

Listed (30)

section	entries
Community SDKs	PHP SDK for Jev
Libraries & Integrations	Truffler, Gut, jev-feels
Agent Integrations (MCP & Skills)	System One Connector, quicksilver, dsh-jev-interceptor, Gatekeeper, jev-harness, Decision-Native Agent Runtime
Coding Agents & Developer Tools	Keel, mu, ESF, jevmem, Jev Code Reviewer
Browser & Computer Use	Jev GUI Delegate, Flick, dejevu
Apps & Browser Extensions	slop-filter, JevBystander, Paper Radar
Evaluation & Benchmarks	jeval
Open Models & Reproductions	vllm-jev, this-that-model, tinyjev, Kev
Games & Real-Time Demos	Jev Plays Pokémon Red, wc3env
Finance	beebots
Articles & Analysis	Jev Cookbook (Chinese)

The one worth checking carefully was TypeSafeAI/jev-harness. The organisation name reads like the vendor's, so we checked before listing: TypeSafeAI is "TypeSafe Community", created 2026-09-18 — three days after Jev launched — with the blog jev.works and no verification. The official organisation is typesafe-ai, created 2024-05-28, which publishes the SDKs. To the project's credit its README says so itself, under "Repository provenance": the community organisation is independent of the official team and this is not an official SDK or an endorsed runtime. The entry carries that distinction, because the URL alone would mislead.

Four more worth naming:

• System One Connector (306★) is the largest addition: one MCP connector giving Claude Code, Claude Desktop, Codex, Hermes and pi an evaluate tool, pointed either at Jev or at an open System One model you host.
• this-that-model ships a 1.9B typed-decision model with an arXiv paper behind it and a /v1/systemone endpoint.
• dejevu's pitch is "no Jev needed" — it runs browser agents off any open model — and it is listed because --backend typesafe makes Jev the chooser, which is the honest way to offer both.
• Decision-Native Agent Runtime is the second repository from the author of jev-agent-design-with-topk-logits-choices, listed since the 25th. The description says it is the same research line with an extended option space, so the pair does not read as two unrelated projects.

Stats regenerated with trending.py's own entry_count: 336 → 366 entries. awesome-lint clean apart from the two pre-existing spell-check warnings on jevql and jevframe.

🤖 Generated with Claude Code

https://claude.ai/code/session_013MQewmnosW2tdUbmHCnryG