Hermes: personal AI agent with a subagent team

Hermes: personal AI agent with a subagent team

2026

Hermes is a personal AI agent: an analyst, an assistant and a task coordinator in one place. Not a chat that just replies with text, but a working assistant that checks, searches, analyzes, writes, runs tools and takes a task all the way to the result.

Hermes: personal AI agent with a subagent team banner 1

The pain it solves

A regular chatbot answers and forgets. Real work needs someone to finish the job: gather data, verify facts, prepare a document, schedule a task. Keeping a separate specialist or a dozen tools for each role is slow and expensive.

Who it's for

Founders, managers and specialists who want one reliable AI assistant instead of scattered tools: a personal assistant, a business analyst, a technical specialist and a content helper at once.

How it works

Hermes breaks a task down and plugs in the right role. Analytics: research, comparison, hypothesis checks, plain-language reports. Documents: specs, guides, README files, decks, commercial copy. Tech: reading and fixing code, working with sites, APIs and logs. Marketing: posts, scripts, competitor analysis. Media: image analysis and generation, visual concepts.

Skills and subagents

Skills are ready-made professional abilities for specific jobs; Hermes has 88 of them across 21 areas of work: browser and websites, YouTube and social media, PDFs and spreadsheets, GitHub and code, email, Notion, Google Docs, Airtable, image generation, market research, cron jobs and integrations. Hermes splits a big task between subagents: one searches, another checks code, a third analyzes competitors, a fourth writes copy — and the results merge into one deliverable.

Hermes: personal AI agent with a subagent team banner 2

Under the hood

An agent platform with model routing via LiteLLM on Google Vertex AI and multi-provider adapters. I designed and built the whole system: from subagent and skill orchestration to finished documents.

Proven in practice

Hermes grew out of an AI marketing analyst: the task history counts 200+ breakdowns, audits and documents, and a typical deep audit takes about 45 minutes. Real clients: a dental clinic chain, a B2B crane manufacturer, an ERP product. Each got an audit, market research and a ready strategy.

Python
LLM
AI agents
subagents
LiteLLM
Vertex AI
GitHub
Telegram
RU