Hermes Agent

Hermes Agent already runs where you talk to it: Telegram, Discord, Slack, WhatsApp, Signal, email, or the CLI. The Apify plugin hands it thousands of ready-to-run Actors from Apify Store, and lets the agent pick, run, and read them on its own.

Hermes Agent is an open-source autonomous agent from Nous Research. It keeps memory across sessions, schedules its own tasks in plain language, and runs on your machine, in Docker, or over SSH. With the Apify plugin, your agent searches Apify Store, reads an Actor’s input schema, starts runs, and collects structured results from a sentence you type.

You install one Python package. The agent decides which Actor fits the request.

• Find the right Actor from a sentence. Ask for "an Actor that gets Instagram profiles" and apify_discover searches Apify Store and returns candidates with their IDs.
• Check inputs before running. The same tool fetches an Actor's input schema and README, so the agent fills in parameters from the schema instead of guessing.
• Start up to ten Actor runs in one call. apify_start fires a batch and hands back run references immediately. The agent keeps working while the runs finish.
• Collect structured results when they land. apify_collect polls those references and returns dataset items once every run reports done.
• Cover the sources your work actually needs. Google Search, Google Maps, TikTok, YouTube, e-commerce sites, and thousands more.

1. Install and enable the plugin. You will need Hermes Agent 0.18.1 or later and Python 3.11 to 3.13.

pip install apify-hermes-agent-plugin
hermes plugins enable apify

2. Save your Apify API token

hermes apify-setup prompts for the token, writes it to ~/.hermes/.env, and enables the apify toolset. Pass --token to skip the prompt.

3. Ask

"Find B2B software companies in Berlin that are hiring sales reps, pull their websites, and draft a one-line opener for each." The agent discovers the right Actors, batches the runs, collects the datasets, and comes back with the list and the openers.

Tool	What it does
apify_discover	Search Apify Store by keyword, or fetch an Actor's input schema and README by ID
apify_start	Start up to 10 Actor runs in one call and return run references
apify_collect	Poll run statuses and return completed dataset results

The three tools work asynchronously: discover, start, collect, then re-call apify_collect until all_done is true. Returned items arrive wrapped as untrusted external content and capped at 50,000 characters per run, so the agent can tell scraped text apart from your instructions.

Lead generation and enrichment. Ask for dental clinics in Austin with a website and no online booking. Hermes Agent runs Google Maps Scraper , crawls each site with Website Content Crawler , and hands back a qualified list with a reason per lead.

Social media analytics. Track how a campaign hashtag performs across TikTok and YouTube. The agent batches the scrapers in one apify_start call, collects posts and view counts, and reports reach and top creators in your chat.

Competitor and price monitoring. Schedule a daily run over competitor product pages with an e-commerce scraper from Apify Store. The agent diffs today's prices against yesterday's and messages you only when something moved.

Market research on demand. "Give me everything published about agentic payments this month." The agent chains Google Search Results Scraper  with Website Content Crawler and returns a brief with sources you can check.

• Runs where you are - Telegram, Discord, Slack, WhatsApp, Signal, email, and the CLI, with one memory across all of them
• Learns as it goes - persistent memory and auto-generated skills carry context between sessions
• Multi-model - runs on Nous Portal, OpenRouter, OpenAI, or any compatible endpoint
• Open source - MIT-licensed, self-hostable on local, Docker, SSH, Singularity, or Modal backends