A Hermes Agent n8n integration works in both directions: Hermes can drive n8n through the official MCP catalog entry, and n8n can call Hermes through a webhook route or the API server.
Get Your Backlink Gameplan →Try Polar Rank Free for 3 Days →
Pick the direction first, because the setup is completely different for each.
If you want agent workflows like these pointed at your SEO and lead flow, get your Backlink Gameplan with my team, or try Polar Rank free for 3 days.
Everything here comes from the Hermes MCP docs, the webhook docs, the API server docs and n8n's MCP server guide.
The three connections
| Direction | Mechanism | Auth | Use it when |
|---|---|---|---|
| Hermes drives n8n | n8n-official MCP catalog entry | Browser OAuth to your n8n instance | The agent should search, run or build n8n workflows |
| n8n triggers Hermes | Webhook route on the Hermes gateway | HMAC secret per route | A workflow step should start an agent run |
| n8n asks Hermes a question | API server, OpenAI-compatible | Bearer key | A workflow needs an answer back in the same execution |
A common pattern is to let n8n run the fixed, repeatable pipeline and let Hermes handle the steps that need judgement.
Connection 1: the n8n-official MCP entry
The n8n-official entry in the Hermes MCP catalog connects straight to your n8n Cloud or self-hosted instance over HTTP.
It uses browser OAuth, so there is no local bridge and no n8n API key to manage.
Steps
- Ask an n8n owner or admin to enable Settings then Instance-level MCP.
- Open Connect and copy the full Server URL ending in /mcp-server/http, not the editor URL. Older n8n versions show it on the MCP settings page.
- Install the entry and paste the URL when prompted:
hermes mcp install n8n-official
- Complete browser OAuth. If you need to redo it:
hermes mcp login n8n-official
- Review the tools it exposes, then start a new session or reload:
hermes mcp configure n8n-official
/reload-mcp
The Hermes backend must be able to reach that URL, which matters if n8n sits on a private network.
What the agent can then do in n8n
n8n's docs list the actions: search your workflows, interact with workflows marked as available in MCP, trigger and test exposed workflows, and create and edit workflows and data tables.
Building and editing workflows is available from n8n 2.13.0.
Instance-level access does not expose every workflow; you enable the ones the agent may see.
Get Your Backlink Gameplan →Try Polar Rank Free for 3 Days →
n8n also has an MCP Server Trigger node that exposes tools from a single workflow, which suits a narrowly scoped agent.
Tool selection matters here
When an MCP entry installs, Hermes probes the server and shows a checklist of its tools to toggle.
Some n8n tools modify or run workflows, so untick anything the agent does not need.
You can change the selection later with hermes mcp configure.
The retired n8n bridge
The older third-party n8n bridge is no longer available for catalog installation.
Existing installs keep their config, credentials and tool selections and still load as custom MCP servers.
It was not migrated to the official entry, and the two live side by side without replacing each other.
If a tutorial tells you to install an "n8n" bridge from the catalog, it predates this change.
🔥 Want the full Hermes plus n8n build on video? 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 running agent workflows. → Join the AI Profit Boardroom
Connection 2: n8n triggers Hermes through a webhook route
The Hermes webhook adapter runs an HTTP server that accepts POSTs, validates signatures and turns each event into an agent run.
Enable the adapter
Run hermes gateway setup and follow the webhook prompts, or set env vars:
WEBHOOK_ENABLED=true
WEBHOOK_PORT=8644
WEBHOOK_SECRET=your-global-secret
Start the gateway and check it:
curl http://localhost:8644/health
The expected reply is a small JSON status saying ok with the platform set to webhook.
Define a route
Routes live under platforms.webhook.extra.routes in config.yaml, or you create them with hermes webhook subscribe.
Point n8n's HTTP Request node at http://your-server:8644/webhooks/
Route options worth knowing
| Option | What it does |
|---|---|
| secret | Required HMAC secret; falls back to the global one; INSECURE_NO_AUTH is for testing only |
| signature | Pins one scheme, such as generic, github or gitlab |
| prompt | Template with dot-notation access to payload fields |
| filters | Declarative checks; non-matches return ignored with HTTP 200 |
| script | A transform script in ~/.hermes/scripts/ that receives the payload on stdin |
| skills | Skill names to load for the run |
| toolsets | Replaces the webhook toolset for this route only |
| deliver | Where the result goes, such as telegram, slack, email or homeassistant |
| deliver_only | Skips the agent and delivers the rendered prompt directly |
| cron_job | Fires an existing cron job instead of a fresh session |
| coalesce | Debounces rapid events on the same entity into one run |
Payload fields are untrusted input, so treat anything templated into the prompt as data, not instructions.
Signing requests from n8n
The generic scheme expects a signature header and a timestamp header computed from the shared secret.
Compute it in an n8n Code node before the HTTP Request, or start with the test-only bypass on a local network and switch to signing before going live.
Connection 3: n8n asks Hermes through the API server
The API server exposes Hermes as an OpenAI-compatible endpoint, so a workflow can send a prompt and get the agent's final answer in the same execution.
API_SERVER_ENABLED=true
API_SERVER_KEY=change-me-local-dev
Start the gateway and it listens on http://127.0.0.1:8642.
From n8n, send a POST to /v1/chat/completions with a Bearer header and model set to hermes-agent.
curl http://localhost:8642/v1/chat/completions \
-H "Authorization: Bearer change-me-local-dev" \
-H "Content-Type: application/json" \
-d '{"model": "hermes-agent", "messages": [{"role": "user", "content": "Hello!"}]}'
The agent runs with its full toolset, including terminal, files, web search, memory and skills, before it answers.
For longer jobs, /v1/runs and /v1/responses keep a conversation across requests using the X-Hermes-Session-Key header.
A run that hits its iteration budget still returns the agent's wrap-up, flagged with finish_reason length.
📞 Workflows humming? Now point them at revenue. Get your Backlink Gameplan and my team will show you the SEO and AI search moves worth automating first, or try Polar Rank free for 3 days. Backlink Gameplan or try Polar Rank free for 3 days
Which direction for which job
| Job | Best connection |
|---|---|
| "Run the weekly report workflow and tell me what failed" | MCP: Hermes drives n8n |
| "When a form is submitted, have the agent draft a reply" | Webhook route with deliver set to email or slack |
| "Classify this ticket and return a label to the workflow" | API server, answer returned in the same execution |
| "Every morning, summarise yesterday's executions" | Cron job in Hermes calling n8n over MCP |
Security checklist
- Untick n8n MCP tools that delete or edit workflows unless you need them.
- Never run a public webhook route without a secret.
- Keep the API server on localhost unless n8n runs elsewhere, and set a strong API_SERVER_KEY.
- Remember that payload fields in prompts are untrusted.
- Keep one gateway per profile to avoid duplicate runs.
Troubleshooting
- MCP install fails to connect: you pasted the editor URL instead of the /mcp-server/http Server URL, or Hermes cannot reach the instance.
- OAuth loop: rerun hermes mcp login n8n-official or use Authorize in the dashboard.
- Workflow missing from the agent's list: it is not enabled for MCP access in n8n.
- Webhook returns ignored: a filter did not match, or the event type is not in the route's events list.
- Webhook returns an auth error: the signature or timestamp does not match the secret.
- API server 401: the Bearer key does not match API_SERVER_KEY.
Worked build: nightly failure digest
This build uses the MCP direction plus a Hermes cron job, and it is a good first project because it only reads.
- Mark your production workflows as available in MCP, and leave only the search and execution-listing tools enabled in hermes mcp configure.
- Create a cron job in Hermes that runs at 7am and asks the agent to list any n8n executions that failed in the last 24 hours.
- Ask it to group failures by workflow, quote the error line, and suggest the most likely cause in one sentence each.
- Set the job's delivery to your home channel so the digest lands in Telegram or Slack before you start work.
After a week you will know which workflows break most, which is the list to fix first.
Because the job never edits anything, a wrong guess costs you nothing but a line in a message.
Worked build: lead triage with a signed route
This build uses the webhook direction and keeps a human in the loop.
- In n8n, a form trigger receives the enquiry and stores it in your CRM as normal.
- A Code node builds the generic signature and timestamp headers from the route secret.
- An HTTP Request node posts name, company and message to the n8n-lead route on port 8644.
- The route prompt asks Hermes to score the lead against your ideal customer notes in memory and draft a reply.
- The route delivers the score and draft to a private Slack channel, where someone approves or edits it.
The route's toolsets list is narrowed to memory and web search only, so a hostile message cannot reach your terminal.
Worked build: in-flow classification
This build uses the API server and returns a value n8n can branch on.
- A support email arrives through n8n's email trigger.
- An HTTP Request node calls /v1/chat/completions on port 8642 with the email body and a system instruction to reply with one word: billing, bug, sales or other.
- A Switch node routes the email to the right queue based on that word.
- Anything the agent labels other goes to a human inbox.
Keep the instruction strict and short so the answer is easy for the Switch node to read.
Sizing and hosting notes
Hermes and n8n can run on the same server, which keeps the API server on localhost and avoids exposing it.
If n8n runs in the cloud and Hermes runs at home, use a private network such as a mesh VPN rather than opening ports.
The webhook adapter's rate limiting and body size limits protect the gateway from floods; keep the defaults until you have a reason to change them.
Long agent runs tie up an n8n execution when you use the API server, so set a node timeout that matches the slowest job you expect.
Operator habits that prevent surprises
Name routes after the n8n workflow that calls them, so logs map back to their source at a glance.
Keep a short README in your Hermes profile that lists every route, its secret's location and its delivery target.
Review the n8n MCP tool selection whenever n8n upgrades, because new versions can add tools.
Test each route with a signed request after changing secrets, before the next real event arrives.
Where to go next
The automation blueprints show cron, GitHub and API-call triggers you can mirror with n8n.
The MCP how-to guide covers filtering tools in more depth, and the n8n Webhook node docs cover the receiving side if Hermes delivers back into n8n.
Members of the AI Profit Boardroom share their Hermes and n8n workflow builds each week.
Get Your Backlink Gameplan →Try Polar Rank Free for 3 Days →
FAQ: Hermes Agent n8n Integration
What is the official Hermes Agent n8n integration?
The n8n-official MCP catalog entry. It connects to n8n Cloud or self-hosted over HTTP with browser OAuth, with no local bridge or n8n API key.
Which URL does Hermes need from n8n?
The full Server URL ending in /mcp-server/http from Settings, Instance-level MCP, Connect. The editor URL will not work.
How does n8n trigger a Hermes agent run?
Send a POST from the HTTP Request node to a Hermes webhook route at port 8644, signed with the route secret.
Can n8n get an answer back from Hermes in one step?
Yes. Enable the API server and call /v1/chat/completions on port 8642 with a Bearer key and model hermes-agent.
What happened to the old n8n bridge?
It is no longer available for catalog installation. Existing installs keep working as custom MCP servers.
📺 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
Install the n8n-official entry today, untick the tools you do not need and run one workflow from chat.
Then get your Backlink Gameplan, and keep this page as your Hermes Agent n8n integration reference, or try Polar Rank free for 3 days.











