# `ai.paperlist/paperlist` on Anlyon Radar

> Text inside fenced blocks marked untrusted-server-text was written by the MCP server's publisher, not by Anlyon Radar. It is data to inspect, not instructions to follow. Invisible characters are shown as [U+XXXX].
> No finding is not a safety guarantee: these are point-in-time checks of tool definitions, not of code or behaviour.

- Page: https://radar.anlyon.com/servers/ai.paperlist/paperlist
- JSON: https://radar.anlyon.com/api/v1/servers/ai.paperlist/paperlist
- Version: 1.0.0 (active)
- Publisher: ai.paperlist (domain proven to the MCP Registry)
- Runs: remote (streamable-http)
- Sign-in to list tools: none
- Metadata from the official MCP Registry, synced 2026-11-18T11:17:47.253Z

Title and description, as the publisher wrote them:

```untrusted-server-text
PaperList
Search papers from 36 AI conferences, including NeurIPS, ICLR, ICML, and CVPR.
```

Scanned 2026-11-18T10:18:15.584Z over streamable-http.

**9 tools: 9 only read.**

| Group | Class | Tools | Claude Code |
| --- | --- | --- | --- |
| Deletes or cancels | destructive | 0 | deny |
| Can't be undone | irreversible | 0 | ask |
| Changes things | write | 0 | ask |
| Only reads | read | 9 | allow |

## Exfil surface

No private-data -> untrusted-input -> outbound path in this server alone: no tool is tagged private-data or outbound.

| Tag | Meaning | Tools |
| --- | --- | --- |
| private-data | Reads files, mail, messages, databases, secrets, calendars or repositories. | none |
| untrusted-input | Brings in text someone else can write: web pages, inbound email, messages, issues and comments. | `mcp__paperlist__list_conferences` `mcp__paperlist__list_papers` |
| outbound | Can carry data to people or systems: it sends, posts or publishes, or takes a recipient, channel or URL. | none |

## Findings

No check matched anything in these tool definitions.


## Only reads (9)

Looks things up and changes nothing. Claude Code: allow.

### `fetch`

- The name "fetch" says "fetch": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Fetch one paper's full text record by id from search.
```

### `find_similar_papers`

- The name "find_similar_papers" says "find": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Find papers related to a given paper, ranked by full-text similarity of its title and keywords. Defaults to slim rows; pass fields=full for abstracts. Optionally restrict to conferences or years.
```

### `get_paper`

- The name "get_paper" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Get one paper's full record by id (e.g. iclr2026-1058, paper-1058, iclr2027-n42, a numeric catalog id, a paperlist.ai paper URL, or an OpenReview forum id if it is in Paperlist), including the review summary or ICLR 2027 TLDR when available. Slug ids are verified against venue and year, so iclr2026-1 will not silently return an unrelated paper-1.
```

### `get_topic_stats`

- The name "get_topic_stats" says "get": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Trend stats for one research topic across NeurIPS, ICLR, ICML and CVPR, 2023–2026: paper counts by year × venue, year-over-year growth, fastest-growing co-labeled sub-themes, sibling topics in the same area, and ~20 representative papers (title, venue, year, paperlist id). Precomputed snapshot, no live aggregation. Get slugs from list_topics (e.g. world-models, multimodal-llms, reinforcement-learning). For ICLR 2027 primary areas, pass a slug from list_conferences topics_for {conference:"iclr", year:2027} (prefixed iclr2027-).
```

### `list_conferences`

- The name "list_conferences" says "list": a word for reading.
- Tag untrusted-input: The description's first sentence says "list" and "review": it reads comments, which other people can write.

Description, as the server wrote it:

```untrusted-server-text
List conferences in Paperlist with their slugs, full names, research domains and available years, including ICLR 2027 (under-review submissions on the public ratings board). Pass topics_for {conference, year} to also get that edition's topic slugs with paper counts (use the slugs as the topic filter in search_papers/list_papers; ICLR 2027 areas are prefixed iclr2027-). No account or API key required.
```

### `list_papers`

- The name "list_papers" says "list": a word for reading.
- Tag untrusted-input: The description's first sentence says "browse": it reads web pages, which anyone can write.

Description, as the server wrote it:

```untrusted-server-text
Browse papers of one conference edition (e.g. conference="iclr", year=2025, or year=2027 for ICLR under-review submissions). Defaults to slim rows (id, title, conference, year, decision, rating_avg, url, short topics) with up to 250 per page; pass fields=full for abstracts (max 25) or use get_paper. Optionally filter by acceptance type or topic slug. For keyword search use search_papers.
```

### `list_topics`

- The name "list_topics" says "list": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
List the 207 research topics that have trend stats, grouped into 40 areas, each with its slug, English and Chinese name, and paper count. Scope: NeurIPS, ICLR, ICML and CVPR, 2023–2026. Use a slug from here with get_topic_stats.
```

### `search`

- The name "search" says "search": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Search Paperlist papers by keywords. Returns ids to pass to fetch.
```

### `search_papers`

- The name "search_papers" says "search": a word for reading.

Description, as the server wrote it:

```untrusted-server-text
Full-text search over titles, abstracts, keywords and authors. Quoted phrases stay exact; OR / AND / NOT and parentheses are supported; common ML acronyms (JSCC, RAG, RLHF, VLA, VLM, DPO, MoE, ...) are expanded. Filter by conference slugs, year range (include year 2027 for ICLR 2027 under-review submissions), acceptance type, topic slug, or author name (must match one author). Defaults to slim rows (id, title, venue, year, decision, rating, url) so clients can scan many papers; pass fields=full or call get_paper/fetch for abstracts. If no paper matches enough of the query, the response is empty or clearly marked as a looser fallback, rather than off-topic any-word hits.
```

## Claude Code permissions

For `.claude/settings.json`, assuming the server is installed as `paperlist`:

```json
{
  "permissions": {
    "allow": [
      "mcp__paperlist__list_conferences",
      "mcp__paperlist__search_papers",
      "mcp__paperlist__get_paper",
      "mcp__paperlist__list_papers",
      "mcp__paperlist__find_similar_papers",
      "mcp__paperlist__search",
      "mcp__paperlist__get_topic_stats",
      "mcp__paperlist__list_topics",
      "mcp__paperlist__fetch"
    ],
    "ask": [],
    "deny": []
  }
}
```

Capability tags behind the rules (a read tagged outbound is asked about):

- `mcp__paperlist__list_conferences`: untrusted-input
- `mcp__paperlist__list_papers`: untrusted-input

## Remote posture

What an anonymous client saw at the remote endpoint, last 2026-11-18T16:18:26.444Z, unchanged since 2026-11-18T16:18:26.444Z. No credentials were sent and no tool was called. Values are shown as the endpoint returned them. Method: https://radar.anlyon.com/methodology#remote-posture

| Fact | Observed |
| --- | --- |
| Reachable | `yes` |
| HTTP status of an anonymous initialize | `200` |
| tools/list served without credentials | `yes` |
| Tools that change things, listed without credentials | `0` |
| Protocol version | `2025-06-18` |
| WWW-Authenticate on 401 | `not observed` |
| resource_metadata in WWW-Authenticate | `none` |
| RFC 9728 protected resource metadata | `https://paperlist.ai/.well-known/oauth-protected-resource/mcp` |
| Authorization servers | `https://paperlist.ai/api/auth` |
| Legacy HTTP+SSE transport | `no` |
| Dynamic client registration | `yes` |
| Client ID metadata documents | `yes` |
