Turn an n8n workflow result into a link
Point n8n's MCP Client Tool at https://mintpage.ai/mcp (HTTP Streamable) and an AI Agent node can turn any workflow result into a live, shareable link.
An n8n workflow ends with a result: a weekly report, a digest of support tickets, a list of leads, a summary an AI Agent wrote. Sending it on usually means pasting into an email, a Slack message or a Google Doc. MintPage adds a publish_page tool to n8n's MCP Client Tool node, so the same result becomes a public URL in one step, and update_page keeps that URL current on the next run.
Connect n8n
MintPage is a remote MCP server over plain HTTP, which is exactly what the MCP Client Tool node expects. In the editor:
- Add an AI Agent node with a chat model attached.
- On the agent's Tool connector, add MCP Client Tool.
- Fill it in:
- Endpoint:
https://mintpage.ai/mcp - Server Transport:
HTTP Streamable - Authentication:
Bearer Auth, with a Bearer Auth credential whose token is your MintPage key (mp_…) - Tools to Include:
All, orSelectedwith justpublish_pageandupdate_pageif you want a smaller surface
- Endpoint:
- Run the node once: it lists the server's tools, and the agent can now call them.
Then tell the agent what to do in its prompt: "Write the weekly summary from the items below, publish it as a page titled 'Week 45 — support digest', and return the URL." The agent calls publish_page and the URL comes back in its output, ready for a Slack or email node downstream.
For guest mode, set Authentication to None: no key, no account, pages last 24 hours. A free key from /start makes them last 7 days and unlocks stats; Pro ($5/month or $39/year) makes them permanent with custom slugs.
Version note. The Endpoint and Server Transport fields appear on MCP Client Tool node version 1.1 and later; HTTP Streamable is the default from 1.2. On n8n releases older than that, the node shows a single SSE Endpoint field and uses the older transport, which MintPage does not serve, so update n8n (the MCP nodes shipped in 1.88) if you only see that field.
Without an LLM in the loop
If the page content is already assembled by earlier nodes (a Code node, a Set node, an HTML template), you do not need an agent at all. Call the server from an HTTP Request node:
- Method:
POST, URL:https://mintpage.ai/mcp - Headers:
content-type: application/json,accept: application/json, text/event-stream, andAuthorization: Bearer mp_your_key(omit for guest mode) - Body (JSON):
{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"publish_page","arguments":{"title":"{{ $json.title }}","content":"{{ $json.markdown }}"}}}
The response carries url, page_id, slug and expires_at. Keep page_id in your workflow's static data (or a sheet) and send update_page with it on the next run, so one URL stays the living copy instead of a new link every time.
Good uses
- Scheduled reports. A Schedule Trigger, a few data nodes, an agent that writes the narrative, and one published page per week, each with a stable link you can pin in a channel.
- Living dashboards from spreadsheets. Read a Google Sheet, render a Markdown table,
update_pagethe same URL every hour. - Form and ticket digests. Collect the day's form submissions or support tickets and publish a readable summary for the people who should not have to open n8n.
- Hand-offs between agents. One workflow's agent publishes a plan; a second workflow (or a person) reads it at the link. The page is the shared memory.
Example call
{"name":"publish_page","arguments":{"title":"Week 45 — support digest","content":"# Support digest, week 45\n\n| Area | Tickets | Trend |\n|---|---|---|\n| Billing | 14 | down |\n| Login | 9 | flat |\n| Exports | 6 | up |\n\n**Action:** the export failures all trace to one customer's 40 MB CSV.","theme":"paper","byline":"Published by an n8n workflow"}}
What stays safe
Pages are public to anyone with the link, and the tool description tells the agent never to publish secrets, so keep API keys and personal data out of the content you hand it. HTML is sanitized (no scripts, forms or iframes) and served with a strict Content-Security-Policy. Store the MintPage key as an n8n credential, never as a literal in a node parameter, and delete a page any time with delete_page.
FAQ
Does n8n support remote HTTP MCP servers? Yes. The MCP Client Tool node connects to any MCP server over HTTP Streamable; MintPage is one URL plus an optional bearer header, with no OAuth dance required (OAuth2 is also offered if you prefer it).
The node shows no tools. Check that Server Transport is HTTP Streamable, that the endpoint is exactly https://mintpage.ai/mcp, and that the credential's token starts with mp_. On older n8n versions that only show SSE Endpoint, update n8n.
Can the workflow update a page instead of creating a new one? Yes: update_page with the page_id from the first publish replaces the content and keeps the URL. That is how a living report works.
Is there a size limit? Guest pages up to 250 KB, free keys 1 MB, Pro 2 MB. Markdown (tables, task lists, code fences) and sanitized HTML are both accepted.
Which tools does the server expose? publish_page, update_page, list_pages, delete_page, pause_page, revoke_link, get_page_stats, get_account, plus image tools (create_image, add_image) and live-session comments on Pro. The agent only needs the first two for most workflows.
Other ways in
The same server works from agent reports, ChatGPT, Claude.ai & Desktop, Claude Code, Codex CLI, Cursor, Gemini CLI, the Markdown API, VS Code, Windsurf. Full tool reference in the docs; live examples on the showcase.
Try it in one line
No account needed. Pages last 24 hours as a guest, 7 days with a free key, forever on Pro.
claude mcp add --transport http mintpage https://mintpage.ai/mcp