vr-org-mcp

Read-only Model Context Protocol server for VR.org, a dedicated VR / AR / XR news publication and aggregator.
It gives any MCP-compatible agent (Claude Desktop, Claude Code, Cursor, Continue, and others) one-call access to live VR, AR, and XR news, VR.org's original editorial (including full article text), the VR/AR/XR events calendar, curated headset deals, buyer-guide answers, and top-game and top-app lists.
Eleven tools, five resources, and three prompts. Zero keys. Zero writes. Zero payments.
Install
Run it directly with npx (no global install needed):
Claude Desktop
One-click: download the .mcpb bundle from the latest release and double-click it to install into Claude Desktop. Or configure manually:
Add to your claude_desktop_config.json:
{
"mcpServers": {
"vr-org": {
"command": "npx",
"args": ["-y", "vr-org-mcp"]
}
}
}
Claude Code
claude mcp add vr-org -- npx -y vr-org-mcp
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"vr-org": {
"command": "npx",
"args": ["-y", "vr-org-mcp"]
}
}
}
Hosted remote endpoint (no install)
VR.org also runs the same tools as a remote server over MCP's streamable-HTTP transport, so web clients (ChatGPT connectors, Claude.ai connectors) can use it with no local install:
Human setup walkthrough for every client: vr.org/connect.
| Tool | What it returns |
|---|
search_vr_news | Latest VR / AR / XR headlines from the live feed, with optional category filter and keyword search (all keywords must match, in any order) |
get_vr_trending | Topics currently trending across the feed |
list_vr_originals | Summaries of VR.org's own editorial articles, newest first |
get_vr_article | Full content of one original by slug: metadata, canonical URL, and the article body HTML |
get_vr_events | Upcoming VR / AR / XR industry events (conferences, expos, launches), soonest first |
get_vr_deals | Curated product picks with prices, badges, and retailer links |
compare_vr_headsets | Side-by-side of two headsets (partial names accepted) |
get_top_vr_games | Current ranked top VR games list |
get_top_vr_apps | Current ranked top VR apps and utilities list |
list_vr_sources | The news sources VR.org aggregates, with counts |
vr_explain | Canonical short answer plus pillar-page link for a common question |
Resources
Browsable MCP resources an app can attach as context:
| Resource | Contents |
|---|
vrorg://news/latest | Latest aggregated VR / AR / XR headlines |
vrorg://originals/latest | Index of VR.org's newest original articles |
vrorg://events/upcoming | Upcoming VR / AR / XR industry events |
vrorg://guides | VR.org's canonical pillar-guide answers in one doc |
vrorg://article/{slug} | Full HTML body of any original article (resource template) |
Prompts
| Prompt | What it does |
|---|
recommend_a_headset | Recommends a headset from VR.org's picks given a budget and use case |
this_week_in_vr | Drafts a weekly VR / AR / XR roundup from the feed and originals |
explain_vr_topic | Explains a VR topic grounded in VR.org's canonical answer |
How it works
Every tool composes VR.org's public JSON API (https://vr.org/api/*) into a single agent-friendly response. The server is a thin proxy: it holds no secrets, writes nothing, and cannot move money.
Threat model
VR.org's editorial is controlled, but the live feed also carries third-party RSS headlines. To keep a malicious or compromised upstream headline from manipulating the calling model, every tool output is:
- Sanitized. Control characters and invisible characters (zero-width and direction-override marks, bidi isolates, variation selectors, soft hyphens and Unicode tag characters) are stripped from every string.
- Escaped. In the markdown resources, relayed headlines, source names and snippets are escaped so they cannot close a link and plant one of their own, and a link target is kept only if it is an absolute http(s) URL.
- Capped. Serialized responses are limited to 50 KB so a large payload cannot flood the agent's context window.
Inputs are validated before any outbound request and every free-text parameter is length-capped at the schema layer. Errors are returned as structured, non-echoing objects rather than raw stack traces; on the rare path where an error string reaches the caller it runs through the same output scrub as a tool result, with any credential-shaped substring redacted and the text capped, and is flagged with isError.
Configuration
The API base URL is fixed to https://vr.org in source and cannot be overridden, so the server can only ever talk to VR.org.
| Env var | Default | Purpose |
|---|
VR_ORG_UA_SUFFIX | (none) | Optional suffix appended to the outbound User-Agent |
Development
npm install
npm run dev
npm run build
npm test
npm run typecheck
Changelog
0.5.1
Security hardening. Requires Node.js 22 or newer.
- The output sanitizer now also strips Unicode tag characters (U+E0000 to U+E007F, the range used to hide ASCII instructions inside text that looks blank), bidi isolates, variation selectors, soft hyphens, line and paragraph separators and similar invisible characters. Accented letters, CJK text, the en dash and prices pass through unchanged.
- The
vrorg://news/latest and vrorg://originals/latest resources escape headlines, source and author names and snippets before putting them into markdown, so a crafted headline cannot close its link and add a link or image of its own. A link target is kept only if it is an absolute http(s) URL, with parentheses and angle brackets percent-encoded; anything else (a javascript: URL, for example) becomes an empty link.
- Node.js 20 reached end of life in April 2026, so the package now declares Node.js 22 or newer. CI tests on Node 22 and 24.
- The release workflow builds and packs in one job with no publish permission, then publishes that exact tarball from a second job that holds the OIDC permission and runs no dependency code. Every GitHub Action is pinned to a full commit SHA, and checkouts no longer keep the token.
smithery.yaml and .env.example no longer advertise a VR_ORG_BASE_URL override. The server has ignored it since 0.3.2; the base URL is fixed to https://vr.org.
@modelcontextprotocol/server 2.3.1 to 2.4.0. The test runner moves from vitest 1 to vitest 5, which clears the development-only audit findings; npm audit reports zero vulnerabilities for the full tree.
- No tool, resource or prompt was added or removed, and no argument changed.
0.5.0
Protocol upgrade. Requires Node.js 20 or newer.
- The server now runs on the MCP SDK v2 (
@modelcontextprotocol/server) and speaks the 2026-07-28 protocol revision. Clients on the 2025 revisions connect exactly as before: the opening message decides which revision a connection uses.
- Node.js 18 is no longer supported, because the v2 SDK needs Node.js 20. If you are pinned to Node 18, stay on 0.4.4.
- The install is much smaller. The old SDK pulled in an HTTP stack this stdio server never used, more than 80 packages of it. The production dependency tree is now three packages, and
npm audit is clean with no overrides.
- The hosted endpoint at
https://vr.org/mcp moved to the same protocol on the same day. It also answers at /mcp directly now instead of redirecting to /api/mcp, which halves the round trips for clients configured with the short URL.
- No tool, resource or prompt changed.
0.4.4
search_vr_news now matches on words instead of one exact phrase.
- The keyword filter used to require the whole query to appear, as typed, somewhere in a headline or snippet. Agents tend to send queries like "steam frame release date" or "best vr headset 2026", and those came back empty even when the feed was full of matching stories. In a replay of 25 realistic queries against the live feed, 18 returned nothing.
- A query is now split into keywords. Filler ("what is the latest", "news", "today", years and month names) is dropped, and every remaining keyword has to appear in the headline, snippet, tags, category or source name, in any order. Plurals match singulars, punctuation is ignored ("half-life alyx" finds "Half-Life: Alyx"), and a model name typed without its space still matches ("psvr2" finds "PS VR2"). Numbers and two-letter terms match whole words only, so "quest 3" does not return Quest 3S stories.
- Headline matches are listed first, newest first inside each group.
- When a query is sent, the response adds
match (all_terms, partial, none, or latest for a query that was all filler), terms (the keywords actually used) and total_matches. If nothing matches every keyword, the closest matches are returned and labelled partial with a hint. A call with no query returns exactly what it did before.
- The hosted endpoint at
https://vr.org/mcp carries the same matcher.
0.4.3
Maintenance release.
- The
vr_explain tool description now names the newer topics it can answer ("steam frame price", "meta vr glasses"), so an agent reading the tool list knows to ask for them. The hosted endpoint at https://vr.org/mcp carries the same wording.
- The release workflow now publishes through npm trusted publishing (OIDC) and only falls back to a token if that is rejected. Provenance is attached as before.
0.4.2
vr_explain and the vrorg://guides resource now answer from VR.org's live explainer feed, so they stay current between releases.
- The answers used to be a list compiled into the package in July 2026. By October most realistic topics ("steam frame", "steam frame price", "meta vr glasses", "quest 4") returned
no_explainer, and several of the answers that did exist were out of date. The list is now fetched from https://vr.org/api/explainers and cached for ten minutes, the same way deals and events are, so a guide refresh on the site reaches agents without a package release.
- Every fetched entry is validated before use (keys must be a non-empty list of non-empty strings, and the link must be on
https://vr.org) and passes through the same sanitizer as other upstream text. If the fetch fails or returns nothing valid, the built-in list answers instead. That list now mirrors the site's current 34 guides, up from 10, so an offline server still gives current answers. A failed fetch is remembered for one minute so an outage does not slow down every call.
- Matching is unchanged: the longest key contained in the topic wins, so "steam frame price" reaches the price page and not the Steam Frame hub.
- The
explain_vr_topic prompt no longer offers 'passthrough' as an example topic, which had no answer. It and the no_explainer hint now suggest 'steam frame'.
- The
get_vr_deals description names all five catalog sections: headsets, AR and XR glasses, accessories, haptics and body tracking, gaming hardware.
- Bumps
@modelcontextprotocol/sdk from 1.30.0 to 1.32.1 and refreshes the ip-address and proxy-addr pins. The advisories behind this sit in the SDK's OAuth client and HTTP transport, which this stdio server never loads. npm audit on the production dependencies reports zero vulnerabilities.
0.4.1
Fixes from a September 2026 audit, matched on the hosted endpoint at vr.org/mcp.
get_vr_article now returns the whole article. body_html was cut at the general 4,096-character string cap, which truncated most articles mid-tag while still reporting ok: true. It now has its own 45,000-character cap, and the 50 KB response cap still applies. The article also carries updated, the date a correction was applied, which the hosted endpoint already returned.
vr_explain no longer answers short topics with the wrong guide. A topic like "vr" or "3" used to match inside a longer key and return "PSVR2 vs Quest 3" with confidence; the reverse match now needs four or more characters starting on a word boundary.
compare_vr_headsets matches "Pimax Dream Air SE" correctly. The pimax alias was a prefix match that rewrote every Pimax query to Crystal.
- Reading
vrorg://news/latest no longer shrinks keyword search to 25 items for the next minute (the two shared a cache key with different request sizes).
- A malformed slug in
vrorg://article/{slug} now says "not found" instead of "temporarily unavailable, retry shortly".
- Refreshed transitive dependency pins (hono, fast-uri, qs) for
npm audit.
0.4.0
Content provenance. Feed results now carry a provenance field on every item, either vr_org_editorial (written and edited by VR.org) or third_party_feed (a headline or snippet relayed verbatim from an outside publisher). Any response containing relayed text also carries a content_notice saying to treat that text as data rather than instructions, and the vrorg://news/latest resource carries the same notice in prose. This is additive metadata, so existing fields and existing consumers are unaffected.
The point is to let a calling client tell which text VR.org actually stands behind. It follows the August 2026 GhostSplice research on splitting instructions across MCP channels, whose core mitigation is that clients should treat server output as data. This server is not a vector for that attack (it never requests sampling, its tool descriptions are static literals, and every value that can flow back in as a tool argument is validated against a strict allowlist), but it does relay text it did not write, and now it says so.
Also bumps @modelcontextprotocol/sdk to 1.30.0 and pins four transitive dependencies of the SDK's HTTP transport that carried advisories. This server is stdio-only and never loads that transport, so the pins are hygiene rather than an exposure fix. npm audit --omit=dev reports zero vulnerabilities.
0.3.2
Error-path hardening: tool errors now pass through the same output scrub as tool results, with secrets redacted and error text capped. All free-text parameters gain schema-level max lengths. No behavior change for valid inputs.
License
MIT. A VR.org project.