Hermes Agent + Wayfern

Show Hermes a web job once. It keeps the skill.

~/.hermes/config.yaml
OAuth or a token header
Cron to Telegram or Discord

Hermes writes its own skills and runs them on its cron. A skill that reads the web stops working when the site blocks bots.

With Wayfern, the skill calls tools: scrape_page for a page, extract_data for fields, or an operation that turns a public page into a JSON call. Hermes records the calls, and Wayfern does the fetching.

hermes · ~/.hermes ⚕ Hermes Agent

$ hermes mcp login wayfern

✓ wayfern: signed in on wayfern.com

❯ Every Monday, list new events on

library.acme.dev. Save it as a skill.

┊ map_site url: library.acme.dev, search: "events"

6 URLs · credits_charged: 2

┊ scrape_page url: library.acme.dev/events

markdown · credits_charged: 1

┊ extract_data urls: […/events/calendar], prompt: "title, date, room"

9 events · credits_charged: 1

library-events ~/.hermes/skills/library-events/SKILL.md

1. map_site map library.acme.dev for "events"

2. scrape_page read /events as markdown

3. extract_data extract title, date, room

cron 0 9 * * 1 · Mondays 09:00 → Telegram

Setup

Add Wayfern to config.yaml.

Hermes runs on your computer or on a server. It reaches Wayfern over the remote server from either.

Install Hermes

Skip this when Hermes is installed.

bash

curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup

2
Add the server

Or run hermes mcp add wayfern --url https://api.wayfern.com/mcp --auth oauth.
~/.hermes/config.yaml

mcp_servers: wayfern: url: "https://api.wayfern.com/mcp" auth: oauth
3
Sign in

A browser opens on wayfern.com. Sign in and click Allow. In a running chat, /reload-mcp loads the server.
bash

hermes mcp login wayfern hermes mcp test wayfern

Teach it a job

prompt

Every Monday, read the events page of the city library with Wayfern, list the new events, and save the steps as a skill.

Other ways to connect

With a token

For a server with no browser. Use a token with the mcp scope.

~/.hermes/config.yaml

mcp_servers:
  wayfern:
    url: "https://api.wayfern.com/mcp"
    headers:
      Authorization: "Bearer ${WAYFERN_API_TOKEN}"

As an agent skill

Hermes copies the skill to ~/.hermes/skills. It takes effect in new sessions. The skill needs WAYFERN_API_TOKEN in the environment.
bash

hermes skills install https://wayfern.com/skills/wayfern/SKILL.md --name wayfern

Route

Single tool calls make good skills.

Hermes learns from its tool calls. MCP gives it one call per step, each with a known cost, and the OAuth sign-in keeps the token out of the skill file.

Remote MCP OAuth sign-in: Best fit; hermes mcp login wayfern on a computer with a browser.
MCP + token Authorization header: Works; A header with ${WAYFERN_API_TOKEN}. Use it on a server with no browser.
Agent skill SKILL.md: Works; hermes skills install with the SKILL.md URL. The skill calls the REST API with curl.

Example tasks

Jobs Hermes can learn

01

List new events

“Every Monday, read the events page of the city library, list the new events and save the steps as a skill.”

scrape_page at 1 credit a page. Hermes writes the skill to ~/.hermes/skills.
02

Turn a page into an operation

“Learn the search page of the used-car site as a Wayfern operation with make and model as parameters. Then call it each morning for a Volvo V60.”

learn_operation costs 3 credits when it finds candidates. create_operation is free. Each call_operation costs 1 credit.
03

Watch for mentions

“Every 6 hours, search Google for new posts about our product on reddit.com and news.ycombinator.com, and send the links to Telegram.”

search_web with site set bills 2 credits for each provider page.
04

Build a dataset

“Crawl the public recipes section of this site, 100 pages at most, and write title, time and ingredients to recipes.jsonl.”

crawl_site bills 1 credit for each page. crawl_status is free.

Tools

What Hermes Agent can call

The same tools on every route, at the REST API rates. Every result reports credits_charged.

All tool inputs

scrape_page Fetch one page through an anti-detect browser and return clean markdown, HTML, links and metadata. The screenshot, pdf and json formats add +4 credits (once); summary adds +2. 1 credit / page
extract_data Extract structured data from one or more pages with a prompt and an optional JSON Schema. Re-running over pages already fetched is free. 1 credit / uncached source page
map_site List a site's URLs from its sitemaps and page links. 2 credits / 10 URLs
capture_screenshot Capture a clean image of a page: cookie banners and ads removed, lazy images loaded, tall pages captured end to end. Returns a short-lived download URL and an artifact key, not image bytes. Slices of a full page cost nothing extra; a fail_if_contains match is not charged. 5 credits / capture
brand_profile Identity, colours, logos, social links and company metadata for a domain. 10 credits / profile
site_styleguide Colours, typography, spacing, radii, shadows and the design tokens a site declares. 10 credits / site
site_fonts Font families, where each is served from, available weights, file formats and where they are used. 5 credits / site
crawl_site Start a bounded multi-page crawl. Returns a job id immediately and charges nothing up front; pages are billed as the crawl runs. 1 credit / successful page
crawl_status Poll a crawl started with crawl_site and read a page of its results. The pages were billed as they were crawled. Free
search_web Search Google, Bing, and DuckDuckGo. Returns ranked results, ads, provider failures, pagination, and related queries. 2 credits / provider page
list_search_engines List live provider ids and labels. Free

Good to know

Skill review: skills.write_approval holds new skill files until you approve them.
Credits: A cron job spends credits on each run. get_usage reports the balance and is free.