Nerve
Asynchronous System-1 supervisory plugin for Hermes Agent with typed Jev decisions, ranking, verification and an opt-in tool gate.
Overview
Nerve is an asynchronous supervisory layer for Hermes Agent that adds typed System One decisions, ranking and verification. Its Reflex component acts as a watchdog and forecaster, while the main Hermes orchestrator keeps final stop or continue authority. The core plugin is dependency-free; model runtimes run as sidecars. The stable release is v0.3.1, with an opt-in ask-only Action Gate for tools.
A provider-neutral backend seam lets Jev be authoritative while a local Laya or OpenJev model shadows it, logging paired decisions; Laya can also be selected directly, with receipts marked LOCAL_ONLY, and shadow failures never alter the authoritative result. Development builds add a crash-safe Jev/Laya/OpenJev benchmark matrix runner. It is for Hermes users who want a fast second opinion on agent progress and actions without slowing the main loop.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Should the agent continue, retry or hand off for review?(inferred) | — |
| Yes / no | Is this tool action safe to run without asking?(inferred) | yes · no |
| Yes / no | Was this task actually completed?(inferred) | yes · no |
How it works
- Hermes runs its normal loop; Nerve observes asynchronously.
- Reflex asks typed questions via Jev (or Laya/OpenJev).
- Shadow backends log paired decisions without changing results.
- Advice and gate prompts go to the orchestrator, which decides.
Strengths
- Does not block the main agent loop.
- Shadow mode compares open models against Jev safely.
- Dependency-free plugin core.
Caveats
- main is a development line distinct from the stable v0.3.1.
- Final authority stays with Hermes; Nerve is advisory.
- Laya sidecar needs separate setup.
Getting started
Install via the Hermes catalog (pinned to reviewed commits); see docs/DEV15B_LAYA_INTEGRATION.md for the optional Laya sidecar.At a glance
- Type
- Project
- Status
- beta
- Language
- Python
- Runtime
- Hermes Agent plugin, sidecars
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Nerve is a supervisory nervous system for Hermes agents, adding typed System…
wh000wh000/awesome-jev-liveTyped System One decisions, ranking, verification, and an opt-in Hermes tool gate using TypeSafe Jev.
hellogumbo/awesome-jevSupervisory layer for Hermes agents that adds typed decisions, ranking and verification asynchronously, with Jev authoritative and an open model shadowing it.
MrJev/awesome-jev
Frequently asked questions
Does Nerve stop the agent?
No, it is a watchdog; the Hermes orchestrator keeps final stop/continue authority.
Can I avoid TypeSafe?
You can select the local Laya backend, or run Laya/OpenJev in shadow alongside Jev.
What is the Action Gate?
An opt-in, ask-only gate on tool actions added in v0.3.1.
Related projects
Software-factory supervisor where Jev independently judges coding agents' completion, test sufficiency, progress and need for a human.
ProjectbetaPythonHermes plugin that asks Jev to review risky plans, check claims against sources, and judge whether a task is really finished.
PluginbetaHermes Agent skill that turns an intention into a checkable finish line, with shell checks and Jev judging the fuzzy requirements.
Agent skillprototypePythonAuditable Zig terminal agent pairing Codex for coding with Jev typed preflight and postflight checks under deterministic policy gates.
CLIproductionZigClaude Code Stop hook that blocks an unverified done: one four-question Jev call only when files changed and no check passed.
PluginbetaTypeScriptClaude Code plugin that moves CLAUDE.md rules into hooks, denying rule-breaking edits and unasked deploys via Jev probabilities.
ProjectbetaTypeScript