Hermes Labs: The AI Reliability Company (and Nous Research’s Labs Plugins)

Hermes Labs is an independent AI reliability engineering company based in San Francisco. It builds open-source tools and publishes research that help AI agents fail less in production. It is not affiliated with Nous Research, the makers of Hermes Agent. That matters because Nous Research also has a GitHub repo with a similar name — hermes-labs-plugins — for experimental Hermes Agent plugins. An earlier version of this page mixed the two up, and I’ve corrected it. Below I cover Hermes Labs the company first, then Nous’s plugin repo, so you know exactly which is which.

Hermes Labs, the AI reliability company

Hermes Labs works on a problem every business running AI agents hits sooner or later: the agent does something wrong, and nobody knows why. Its view is that not every agent failure is a model failure. Prompts, the way conversation history is trimmed, tool bindings and evaluation code can all change what an agent receives or does. Hermes Labs reproduces those failures, works out which layer the evidence points to, and engineers fixes backed by regression tests.

Its public work includes fixes merged by the maintainers of Microsoft Semantic Kernel, LangChain and DSPy, and a vulnerability-record correction acknowledged by CISA. It also offers paid diagnosis and engineering for teams running agents in production.

Hermes Labs open-source tools

Hermes Labs releases its tools on GitHub. The best known include:

• lintlang — static analysis for agent configs, tool descriptions and system prompts. It catches problems like vague tool descriptions, missing stop conditions and schema gaps before they reach runtime, with no LLM needed.
• little-canary — prompt-injection detection that sends untrusted input to a powerless “canary” model first and checks how it behaves before your main model acts. It’s in the Hermes Agent plugin catalog, listed as independent of Nous Research.
• zer0dex — a local, dual-layer memory pattern for AI agents: a readable markdown index plus semantic search.
• fidelis — local-first agent memory for Claude Code and other agents that returns original passages word for word.
• hermes-jailbench — a repeatable jailbreak regression benchmark for LLM endpoints.

It also publishes research papers on failure modes in language-model systems. If you build agents for clients, lintlang and little-canary are the two I’d look at first.

Nous Research’s Hermes Labs plugin repo

Here’s where the confusion comes from. On 21 October 2026, with Hermes Agent v0.22.4, Nous Research launched a GitHub repo called hermes-labs-plugins. It holds Hermes features Nous decided not to put into the core product, released so anyone can install them. This repo belongs to Nous Research and has nothing to do with Hermes Labs the company.

Nous is clear that these plugins are not supported. There’s no promise of fixes, they sit in the community tier of the plugin catalog, they’re up for adoption, and CI tests them against the Hermes nightly build so breakage is at least visible. Supported plugins live in hermes-official-plugins.

Nous labs plugin #1: Inbox

Install it with hermes plugins install inbox. Then hermes inbox shows everything in flight for your active profile in one read-only view: failed cron runs, stalled or dropped delegations, orphaned background processes, blocked kanban tasks, running work, undelivered results, upcoming jobs and open chat sessions. If you run scheduled jobs and sub-agents, this is the one I’d install first.

Nous labs plugin #2: Workflow Record

Install it with hermes plugins install workflow-record. Then hermes record captures a browser workflow you demonstrate once — clicks, field values and page visits, with passwords masked. Its bundled skill lets /learn turn the recording into a skill Hermes can replay. Show Hermes a task once, and it learns to do it.

Nous labs plugin #3: Telegram Business

Install it with hermes plugins install hermes-telegram-business. It turns Hermes into an observe-with-approval secretary for your Telegram Business account: it drafts replies to customers, and nothing is sent until you approve it.

How to tell them apart

Quick rule: if it’s a company, a research paper or a reliability tool, it’s Hermes Labs. If it’s an unsupported plugin from the makers of Hermes Agent, it’s Nous’s labs repo.

How to install Nous’s labs plugins

• Update Hermes first with hermes update — the labs plugins arrived in v0.22.4.
• Browse the plugin catalog and look for the community tier.
• Run hermes plugins install inbox (or workflow-record, or hermes-telegram-business).
• Keep them current with hermes plugins update.

Test any community plugin on a separate Hermes profile first and keep approvals on. For the full catalog, see my guide to the Hermes Agent plugins marketplace.

My verdict on Hermes Labs

If you build or run AI agents seriously, Hermes Labs the company is worth following for its open-source reliability tools and research — lintlang and little-canary especially. And if you use Hermes Agent, Nous’s separate labs plugins, especially Inbox, are worth a test. Just don’t confuse the two: Hermes Labs is an independent company, not part of Nous Research.

About Julian Goldie

I’m Julian Goldie — founder of Goldie Agency and the AI Profit Boardroom. I test new AI agents almost every day for 432K+ YouTube subscribers and run Hermes, OpenClaw, OpenCode, Claude and Codex side by side in my own Agent OS. Claim your free strategy session or join the AI Profit Boardroom.

FAQ: hermes labs

Is Hermes Labs a company?

Yes. Hermes Labs is an independent AI reliability engineering company based in San Francisco (hermes-labs.ai), and it is not affiliated with Nous Research.

What does Hermes Labs make?

Open-source tools and research for AI agent reliability, such as lintlang (static analysis for agent configs and prompts), little-canary (prompt-injection detection), zer0dex and fidelis (agent memory), plus paid diagnosis and engineering.

Is Hermes Labs part of Nous Research?

No. Nous Research separately runs a repo called hermes-labs-plugins for experimental Hermes Agent plugins, but Hermes Labs the company is independent.

What is hermes-labs-plugins?

Nous Research’s repo of experimental, unsupported Hermes Agent plugins — Inbox, Workflow Record and Telegram Business — launched with Hermes Agent v0.22.4.

How do I install Nous’s labs plugins?

Update Hermes, then run hermes plugins install followed by the plugin name, such as inbox or workflow-record.

Related reading

• Hermes Agent Plugins Marketplace
• Best Hermes Agent Plugins
• Hermes Agent Sandbox

Last updated October 2026. This page is a living guide to hermes labs — the facts here move fast and it is updated as they do.