Hermes Agent for real estate scores 6.3 out of 10 on my scorecard: strong on channels, scheduling and drafting, weak on property data and on stopping outbound messages.
The ratings below are my editorial judgement from the Hermes v0.23.0 docs, the plugin catalog and the Nous-approved MCP catalog, not from benchmark runs.
Each score carries a one-line reason you can check against the linked source.
If you want my team's view on your own agency's setup, book a free SEO strategy session.
The Scorecard: Ten Criteria Rated Out Of 10
| Criterion | Score | Reason from the docs |
|---|---|---|
| Lead intake channels | 8 | Email, WhatsApp, SMS, Telegram, Slack, Teams and more through one gateway. |
| Follow-up scheduling | 8 | Cron jobs with skills attached and delivery to a chat or platform target. |
| Viewing scheduling | 7 | Calendly is a Nous-approved MCP; Google Calendar via the bundled workspace skill. |
| Listing description drafts | 7 | Skills hold your house style; the vision tool reads photos; PDFs are readable. |
| Area research | 7 | A real browser plus web search and extract, but no property data feed. |
| CRM reach | 5 | Attio and Close are catalogued; no real estate CRM has an entry. |
| Outbound message safety | 5 | Approvals cover dangerous commands and untrusted MCP writes, not gateway replies. |
| Property data and portals | 2 | No MLS, Zillow or portal entry; one Japan-only community data plugin. |
| Control of client data | 9 | Self-hosted, with self-hosted model endpoints such as Ollama and vLLM supported. |
| Setup effort | 5 | Mailbox auth headers, Meta business verification and a public HTTPS URL for WhatsApp. |
The ten scores average 6.3, and the low ones are all about the property world's own software rather than about Hermes itself.
Why Channels Score 8
The messaging gateway is one background process that connects Telegram, Discord, Slack, WhatsApp, Signal, SMS, Email, Microsoft Teams, LINE and more.
For agencies, the WhatsApp Business Cloud API path is the one that matters, because it runs on Meta's official API with a dedicated business number.
The email gateway replies in-thread over IMAP and SMTP and has a sender allowlist, which suits an office that forwards enquiries in.
It loses two points because every channel needs its own setup, and WhatsApp brings Meta's 24-hour template rule with it.
Why Scheduling Scores 8 For Follow-Ups And 7 For Viewings
Scheduled tasks accept plain language or cron expressions, can run once or repeat, and can carry one or more skills.
A weekday job that lists quiet leads and drafts a chaser for each is exactly the kind of work they are built for.
The default cron_mode: deny blocks dangerous commands in scheduled runs, which keeps an unattended job from doing anything drastic.
It isn't a 9 because a job still needs a CRM it can read, and that is where the next scores drop.
The Calendly manifest describes scheduling links, events and invitees, served from Calendly's own hosted MCP with OAuth.
That covers handing a buyer the right booking link and checking who has booked.
The bundled google-workspace skill reaches Calendar as well, for offices that live in Google.
It's a 7 because there is no native diary sync with any real estate platform.
🔥 Want the full scorecard turned into a working build? Inside the AI Profit Boardroom you get step-by-step video trainings, the installable Agent OS, four coaching calls a week and 3,400+ members testing agent setups. → Join the AI Profit Boardroom
Why Listing Drafts Score 7
Skills let you write your office's house style once as a SKILL.md and reuse it on every instruction.
Photos sent by email are passed to the vision tool, and PDF attachments become files the agent can read.
The weak spot is accuracy: the draft is only as good as the notes, so a person must check every measurement and feature.
A good skill fixes the order of the description, bans claims you can't back up and tags every measurement for checking.
Write the rule "describe the home, not the buyer" into it, and have your broker confirm the advertising rules where you sell.
Never let image generation alter how a property looks; the vision tool is there to read photos, not to change them.
Why Area Research Scores 7
The browser tool browser_exec can navigate, read, click, type, scrape and take screenshots.
That handles council planning notices, transport and school pages and local news, with sources you can click.
It stops at 7 because a browser reading public pages is not a property data feed, and site terms still apply.
Why CRM Reach Scores 5
Two general sales CRMs are in the Nous-approved MCP catalog: Attio, for records, lists and notes, and Close, for leads, opportunities, calls and emails.
Twenty CRM has a community plugin in the plugin catalog.
That's all.
HubSpot, Follow Up Boss, kvCORE, Lofty and the other agent CRMs have no entry, so many offices would need to move CRM to get this working.
Why Outbound Safety Scores 5
The security page sets approval modes for dangerous shell commands, with smart as the default.
The MCP config reference adds a trust: untrusted setting that makes every write-capable call on that server ask first.
Neither covers a reply that the agent sends through a gateway.
So the only safe pattern for client messages is "draft only" in the job prompt, with a person sending.
Close does earn credit here: its catalog entry switches off the voice-agent tools by default, including one that places real outbound AI phone calls.
Why Property Data Scores 2
There is no MLS, IDX, Zillow, Realtor.com, Redfin or portal entry in either catalog at v0.23.0.
The only real-estate-specific entry is the community jp-reinfolib plugin, which returns quarterly transaction data for one city in Japan.
Signatures score the same way: no DocuSign and no dotloop.
In practice that means the agent never sees a live listing unless someone pastes or forwards it.
Comparable sales, price history and availability all have to come from your own systems, read by a person.
For a property business, that is the single biggest reason the overall score sits in the middle rather than near the top.
📞 Want a straight answer on whether this fits your agency? Book a free SEO strategy session and my team will review how you win instructions and where an agent could help. → Claim your free strategy session
How I Scored It
Every criterion is a job a small agency does each week, from answering enquiries to writing descriptions.
For each one I checked what the v0.23.0 docs, the plugin catalog and the MCP catalog actually contain.
A 9 means the docs cover the job fully with little setup.
A 5 means it works only for some offices or needs a workaround.
A 2 means the integration you would expect is missing.
Nothing here comes from timing runs or test results, and no score is a claim about outcomes for your business.
Three Offices, Three Verdicts
A four-person sales agency on Close: recommended.
- Staff forward portal enquiries to the agent's mailbox.
- Hermes looks up the lead in Close, drafts a reply with a Calendly viewing link and returns it in the thread.
- A weekday job lists quiet leads in the team chat with a draft for each.
- Negotiators send what they approve from their own accounts.
Every piece of this exists in the catalogs, which is why it rates well.
A lettings agency on an all-in-one property CRM: try with care.
- The CRM has no Hermes entry, so the agent works only from forwarded emails and pasted notes.
- It can still draft replies, sort repair requests and run a Friday area brief.
- Follow-up lists need a shared spreadsheet the agent can read through the google-workspace skill.
Useful, but you are maintaining a second record of your leads.
A brokerage that lives inside its MLS: not yet.
- Listings, showings and documents all run through software with no Hermes integration.
- The agent could only draft copy from notes typed into it.
That is too thin to justify the setup for most brokerages today.
Score Changes To Watch For
The catalog is growing fast: the v0.23.0 release notes describe it passing 1,050 entries.
A real estate CRM or a signature tool arriving in either catalog would lift CRM reach and property data straight away.
A documented approval step for gateway replies would lift outbound safety.
Until then, the scores above stand.
Strengths And Limits At A Glance
Strengths
- Self-hosted, so client names and viewing notes stay on hardware you choose.
- Any model, including self-hosted endpoints, per the providers page.
- Separate profiles for sales, lettings or a second office, each with its own memory and skills.
- Read-only CRM lookups run without prompts, while writes can be gated.
Limits
- No property portal, MLS or signature integrations.
- No built-in approval step before a gateway reply goes out.
- WhatsApp needs a Meta Business account, a business number and a public HTTPS URL.
- The community Microsoft 365 plugin requires a Hermes version below 0.22, so it doesn't load on v0.23.
Who Should Use It, And Who Shouldn't
Use it if you run a small office on Attio, Close or Twenty, take most enquiries by email, and want drafts, chasers and area notes prepared for review.
Use it if keeping client data on your own server is a priority.
Skip it if your day runs inside Follow Up Boss, kvCORE or an MLS and you need the agent working inside them.
Skip it if nobody on the team will read drafts before they go out.
For outbound prospecting rather than inbound enquiries, read our Hermes Agent for sales verdict instead.
Final Verdict
Hermes is a capable drafting and scheduling assistant for an agency that can work around missing property integrations.
It scores well where general business tools are enough, and poorly where only real estate software will do.
For a small office already on Attio or Close, it's an easy yes for drafting and reminders.
For an office tied to an MLS or an agent-specific CRM, wait for the catalog to catch up.
Either way, the house rule is the same: the agent drafts, a person sends.
Rate your own office on the same ten rows before you install anything.
If you want help with that, the AI Profit Boardroom runs coaching calls four times a week.
Also On Our Network
- 🌐 The seven-step lead desk build, from profile to weekday call list
- 🌐 Five small-agency use cases and the risks an owner should weigh
- 🌐 Five copy-paste prompts that set up a Property Desk Bot
- 🌐 Which AI models to run for property pages, listing drafts and private notes
- 🌐 The complete Hermes Agent for Real Estate guide on the AI Profit Boardroom blog
FAQ: Hermes Agent For Real Estate Verdict
What is the overall score for Hermes Agent in real estate?
It averages 6.3 out of 10 on my ten criteria. The scores are editorial judgements from the docs and catalogs, not benchmark results.
What is the biggest weakness for estate agents?
Property data. There is no MLS, Zillow or portal integration in either Hermes catalog, so the agent cannot pull or post listings.
Does Hermes ask before it sends a message to a client?
Not through a gateway. Approvals cover dangerous commands and writes on MCP servers marked untrusted, so client messages should stay as drafts that a person sends.
Which part of Hermes scored highest?
Control of client data scored 9, because Hermes is self-hosted and supports self-hosted model endpoints such as Ollama and vLLM.
📺 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 office on these ten rows first, and book a free strategy session if you want a second opinion before you commit to Hermes Agent for real estate.
