Hermes Agent for lead generation earns a strong verdict for finding and qualifying prospects, and a cautious one for anything that touches outreach.
I scored eleven top-of-funnel jobs from the official docs and the v0.23.0 catalog, not from a lab test.
The ratings are my editorial reading of what each feature is built to do and where the docs draw its limits.
If you'd like my team to sanity-check your own lead flow, book a free SEO strategy session.
The Scorecard At A Glance
| Criterion | Score /10 | Why, from the docs |
|---|---|---|
| Qualification rules | 9 | Skills hold your profile, and /learn builds one from notes or a PDF. |
| Scheduled lead lists | 9 | Cron takes plain-language schedules, attaches skills and delivers to chat or files. |
| Web prospect research | 8 | web_search (limit 1-100) and web_extract (up to 5 URLs a call) cover most digging. |
| Interactive pages | 8 | browser_exec drives Chromium, a cloud browser or your own Chrome. |
| Inbound form triage | 8 | Webhook routes verify HMAC signatures and can load skills per route. |
| List building | 8 | Airtable is a Nous-approved MCP; Sheets runs through the bundled skill. |
| Data control | 8 | Local storage, no telemetry, and a fully local model option. |
| Buying-signal discovery | 7 | Read-only x_search plus a community plugin for public forums. |
| CRM handoff | 6 | Attio and Close only; no HubSpot, Pipedrive or Salesforce CRM entry. |
| Cold email compliance help | 6 | One community plugin checks recipients; the duty stays with you. |
| Outreach safety | 5 | No built-in hold on outbound messages, so drafts only. |
Average across the eleven: about 7.5 out of 10.
The spread tells the story better than the average: research and scheduling score high, sending and integrations score low.
Qualification Rules: 9/10
The skills system is the reason this tool suits lead work at all.
You describe a good customer once, the agent saves it as a skill in ~/.hermes/skills/, and every job can load it.
/learn can author that skill from pasted notes, a docs URL or a PDF, which means your sales one-pager becomes the scoring rubric.
It loses a point only because the skill is just as good as the rules you write, and vague rules give vague scores.
The agent can also edit or delete skills, so keep a copy of the profile file in version control or a shared drive.
Ask for a reason with every score and the rubric stays easy to audit: a wrong reason points straight at the rule that needs fixing.
Scheduled Lead Lists: 9/10
The cron feature accepts plain language, so "every weekday at 8am, find five new fits" is a valid job.
Jobs can attach several skills, deliver to the chat they came from or to a file, and pin a specific model.
There's also a script-only mode with no model call at all, useful for a plain export.
I hold back one point because unpinned jobs follow whatever chat model is set when they fire, which can surprise a team.
A safety detail I like: cron runs can't create new cron jobs, so a badly worded prompt can't multiply itself.
Webhook routes can also fire an existing cron job the moment an event lands, so the morning list and the instant form check can share one prompt.
Web Prospect Research: 8/10
The tools reference documents web_search at five results by default, up to 100, with operators like site: working when the backend supports them.
web_extract returns clean text, handles PDFs and takes up to five URLs a call.
That's enough for most small-business prospecting, from local clinics to niche suppliers.
It isn't a contact database, though, and it doesn't pretend to be one.
Every result depends on the backend you choose, such as Exa or Firecrawl, and on what businesses publish about themselves.
That limit is healthy for compliance: the agent works from public pages with a source link, not from a bought list of strangers.
🔥 Want the scoring skill and cron job without the trial and error? Inside the AI Profit Boardroom you get step-by-step video trainings on Hermes, the installable Agent OS, four coaching calls a week and 3,400+ members testing the same workflows. → Join the AI Profit Boardroom
Interactive Pages And Inbound Forms: 8/10 Each
When a directory needs clicks, the browser tool steps in with Hermes-launched Chromium, Lightpanda, a cloud browser or your own Chrome over CDP.
The docs themselves steer simple lookups back to search and extract, which is the right instinct.
On the inbound side, webhook routes validate HMAC signatures, filter payloads, load named skills and deliver to Telegram, Slack, email and more.
The docs also warn that signed payload fields still count as untrusted, which is a mature design choice for public forms.
Neither gets a 9 because both need config-file work most owners won't enjoy.
One plus for the browser: saved logins live in the credential vault and are filled into the page without the model seeing them.
List Building And Data Control: 8/10 Each
Airtable sits in Nous's approved MCP catalog, and Google Sheets is reachable through the bundled google-workspace skill.
Data control is good by default: the FAQ says there's no telemetry and that conversations, memory and skills stay in ~/.hermes/.
Run a local model and the prospect list never leaves the machine.
Each loses two points for setup friction: OAuth for Airtable, a CLI for Sheets, and real hardware for a local model.
Buying-Signal Discovery: 7/10
x_search is read-only public X search that can't post, reply or DM, and it's off until you add xAI credentials.
For forums and social threads, the community RedReplier plugin finds keyword mentions on Reddit, X, Bluesky, Facebook and a popular tech news forum and scores each from 0 to 100.
It's community-tier and runs through RedReplier's hosted service, so it isn't part of Hermes itself.
That reliance on an outside account is why this sits at 7.
For companies you already know by name, the bundled competitor-news-monitor skill watches for material news and returns cited digests, which works as a quieter signal feed.
Hiring signals have one more source: Indeed is in the approved MCP catalog for searching job listings, though it surfaces jobs, not contacts.
CRM Handoff: 6/10
Attio and Close are the two CRMs in the approved catalog, with Calendly alongside for booking.
If you live in HubSpot, Pipedrive or Salesforce, there's no matching entry, and Apollo, Hunter and Clay are absent too.
A CSV export works, but it's a manual step, and that costs points for many teams.
Attio and Close both connect as vendor-hosted remote servers with OAuth sign-in, so no CRM password sits in a config file.
Mark either one trust: untrusted in config.yaml and Hermes asks before each write call, which keeps junk rows out of your pipeline.
Cold Email Compliance Help: 6/10
The community Precleared plugin screens each new recipient against a set of national cold-email rules, CAN-SPAM and CASL among them, and anything that fails lands in Hermes's approval prompt.
That's useful, but it's a third-party check that sends recipient details to its own service.
The real duty stays with you: respect site terms and the anti-spam and privacy law in your market and your prospect's.
📞 Unsure whether your lead list would pass a basic sense check? Book a free SEO strategy session and my team will walk through your sources with you. → Claim the free session
Outreach Safety: 5/10
This is the lowest score, and it's deliberate.
Approvals in the security model cover dangerous commands and write calls to MCP servers you mark untrusted.
Nothing in the docs holds an email or DM for sign-off.
So the honest rating for unattended outreach is low, and the right workflow is "draft only" in every job, with a person pressing send.
How The Scores Play Out In Three Businesses
A roofing contractor wants property managers in two towns: search and extract find them, a skill scores them by portfolio size, and the high marks for research and scheduling carry the whole job. A design studio lives in HubSpot: the research side scores just as well, but the 6 for CRM handoff bites, because every qualified lead needs a manual import. A print shop wants to email local event organizers: the qualifying works, but the 5 for outreach safety means the owner writes and sends each message personally.In all three, the pattern holds: the agent shortens the list and sharpens the notes, and a person makes the contact.
After qualifying, the closing side is a separate job, rated in my Hermes Agent for sales verdict.
How I'd Lift The Three Low Scores
Outreach safety (5): make every delivery target your own chat or inbox, write draft only into the prompt, and leave YOLO andapprovals.mode: off untouched.
CRM handoff (6): keep Airtable or a sheet as the master list and move only the qualified rows into the CRM, with the untrusted flag switched on.
Compliance help (6): write your market's opt-out and consent rules into the profile skill, so the agent drops people before they ever reach a draft.
None of these change the scores, because they're things you do, not things the tool does.
They do turn a cautious setup into a safe one.
Strengths, Limits, Who Should Use It
Strengths: rule-based scoring, plain-language schedules, solid web research, signed inbound webhooks and local data by default. Limits: a short CRM list, no outbound hold, and config-file setup. Use it if you sell to businesses or a clear local niche, can describe a good customer, and want a reliable daily list. Skip it if you need a ready contact database, rely on a CRM outside the catalog, or want an agent that sends for you.The Final Verdict
For research and qualification, Hermes rates highly, because skills and cron fit the shape of the job so closely.
Give it a written profile, a search key and a list, and it does the reading and sorting that small teams rarely have hours for.
For outreach, treat it as a drafting tool and nothing more.
That split is the whole verdict: high marks where the agent reads and sorts, low marks where it would speak for your business.
Build to the high marks, keep a human on the low ones, and the tool earns its place on a small team.
Want to learn the setup with others doing the same? The AI Profit Boardroom has the video walkthroughs.
Also On Our Network
- 🌐 The eight-step prospect desk build, from ICP skill to morning cron list
- 🌐 Four small-business lead-gen use cases and who should hold off
- 🌐 Five copy-paste Agent OS prompts for finding and scoring prospects
- 🌐 Which benchmarked models to pin to each Hermes lead-gen job
- 🌐 The complete Hermes Agent for Lead Generation guide on the AI Profit Boardroom blog
FAQ: Hermes Agent For Lead Generation Verdict
Where does Hermes score lowest for lead generation?
Outreach safety, at 5 out of 10. The docs describe no built-in hold on outbound messages, so every outreach job should be draft only.
Were these scores from a benchmark?
No. They are editorial ratings based on what the official Hermes docs and v0.23.0 catalog say each feature does.
Which CRM works best with Hermes for leads?
Attio or Close, because they are the CRMs in Nous's approved MCP catalog. Other CRMs need a CSV export.
What is Hermes best at in prospecting?
Qualification rules and scheduled lists, both scored 9. Skills hold your customer profile and cron delivers a scored list on a schedule.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
About Julian
I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.
I help business owners scale with AI agents, automation and SEO.
- I've built a 7-figure SEO and link building agency, Goldie Agency, with a team of 70+ SEO experts.
- I've grown a YouTube channel to 400,000+ subscribers.
- I wrote "Link Building Mastery".
- My Udemy courses have taught over 29,000 students.
→ Get my best AI training inside the AI Profit Boardroom
Score your own five criteria against the table, keep outreach in draft, and book a free SEO strategy session if you want a second view on Hermes Agent for lead generation.
