Skip to content
Back to search

pwc-cli

A Command-Line Interface (CLI, Skill and MCP Server to interact with Papers with Code. For agents and humans.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/huggingface/pwc-cli with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/huggingface/pwc-cli'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/huggingface--pwc-cli/badge.svg)](/listings/huggingface--pwc-cli)

HTML

<a href="/listings/huggingface--pwc-cli"><img src="/api/listings/huggingface--pwc-cli/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Prebuilt binary (recommended)

The installer downloads the binary for your platform, verifies its SHA-256 checksum, and installs it as `~/.local/bin/pwc`:

pwc_repo=https://raw.githubusercontent.com/huggingface/pwc-cli/main
curl -LsSf "$pwc_repo/standalone_cli/install.py" | python3

The installer selects the latest release by default. Pass `--version VERSION` after `python3 -` to explicitly pin a release. It also reports whether `~/.local/bin` is on `PATH` and prints a persistent shell instruction when needed.

Prebuilt releases support:

- Linux x86-64 - macOS Intel (x86-64) - macOS Apple Silicon (arm64)

Every release includes checksum files and GitHub build-provenance attestations. See the [latest release](https://github.com/huggingface/pwc-cli/releases/latest) to download and verify an artifact manually.

Install from source

Python 3.10 or newer is required for source installations.

With [`uv`](https://docs.astral.sh/uv/):

uv tool install 'git+https://github.com/huggingface/pwc-cli.git#subdirectory=standalone_cli'

Or with [`pipx`](https://pipx.pypa.io/):

pipx install 'git+https://github.com/huggingface/pwc-cli.git#subdirectory=standalone_cli'

To work from a clone instead:

git clone https://github.com/huggingface/pwc-cli.git
cd pwc-cli
uv tool install ./standalone_cli

Confirm the installation:

pwc version

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 394 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # pwc CLI, Skill, and MCP server
  2. A fast command-line client for exploring and editing the
  3. [Papers with Code](https://paperswithcode.co) catalog, alongside a Skill for
  4. coding agents and a hosted Model Context Protocol server.
  5. ## MCP server
  6. The [`mcp_server`](mcp_server) package exposes an anonymous, read-only,
  7. agent-native interface over Streamable HTTP. It supports MCP `2026-07-28` and
  8. legacy 2025-era clients from one endpoint, with one tool per `pwc` research
  9. command: papers, related work, lineage, tasks, methods, conferences,
  10. organizations, frameworks, benchmarks, and filtered leaderboard evaluations.
  11. A hosted instance runs at `https://paperswithcode.co/mcp`. It needs no API key
  12. or sign-in. Connect your client:
  13. ```bash
  14. # Claude Code
  15. claude mcp add --transport http paperswithcode https://paperswithcode.co/mcp
  16. # Codex
  17. codex mcp add paperswithcode --url https://paperswithcode.co/mcp
  18. ```
  19. For Cursor (`.cursor/mcp.json`) and other clients that read an `mcpServers`
  20. config:
  21. ```json
  22. {
  23. "mcpServers": {
  24. "paperswithcode": {
  25. "url": "https://paperswithcode.co/mcp"
  26. }
  27. }
  28. }
  29. ```
  30. In VS Code (`.vscode/mcp.json`), use `"servers"` with `"type": "http"` and the
  31. same URL. In Claude.ai or Claude Desktop, add it under Settings → Connectors →
  32. Add custom connector.
  33. Run it locally with:
  34. ```bash
  35. uv run --project mcp_server pwc-mcp
  36. ```
  37. See [`mcp_server/README.md`](mcp_server/README.md) for tools, resources,
  38. configuration, security limits, and hosting details.
  39. Let your coding agent use `pwc` to:
  40. - search papers
  41. - browse benchmarks and find state-of-the-art (SOTA) by domain
  42. - read paper metadata and Markdown
  43. - discover recent, trending, and related work
  44. - inspect paper lineage (follow-up work and predecessors)
  45. - explore tasks, methods, conferences, organizations, and frameworks.
  46. Public reads are anonymous: no account, API token, or Python environment is required.
  47. ```bash
  48. pwc search "small vision language models" --limit 3
  49. pwc paper info 2501.01234
  50. ```
  51. It's recommended to also install the Skill so your coding agent knows how to use the CLI.
  52. ## Installation
  53. ### Prebuilt binary (recommended)
  54. The installer downloads the binary for your platform, verifies its SHA-256
  55. checksum, and installs it as `~/.local/bin/pwc`:
  56. ```bash
  57. pwc_repo=https://raw.githubusercontent.com/huggingface/pwc-cli/main
  58. curl -LsSf "$pwc_repo/standalone_cli/install.py" | python3
  59. ```
  60. The installer selects the latest release by default. Pass `--version VERSION`
  61. after `python3 -` to explicitly pin a release. It also reports whether
  62. `~/.local/bin` is on `PATH` and prints a persistent shell instruction when
  63. needed.
  64. Prebuilt releases support:
  65. - Linux x86-64
  66. - macOS Intel (x86-64)
  67. - macOS Apple Silicon (arm64)
  68. Every release includes checksum files and GitHub build-provenance attestations.
  69. See the [latest release](https://github.com/huggingface/pwc-cli/releases/latest)
  70. to download and verify an artifact manually.
  71. ### Install from source
  72. Python 3.10 or newer is required for source installations.
  73. With [`uv`](https://docs.astral.sh/uv/):
  74. ```bash
  75. uv tool install 'git+https://github.com/huggingface/pwc-cli.git#subdirectory=standalone_cli'
  76. ```
  77. Or with [`pipx`](https://pipx.pypa.io/):
  78. ```bash
  79. pipx install 'git+https://github.com/huggingface/pwc-cli.git#subdirectory=standalone_cli'
  80. ```
  81. To work from a clone instead:
  82. ```bash
  83. git clone https://github.com/huggingface/pwc-cli.git
  84. cd pwc-cli
  85. uv tool install ./standalone_cli
  86. ```
  87. Confirm the installation:
  88. ```bash
  89. pwc version
  90. ```
  91. ## Add the CLI Skill
  92. Skills give coding agents the context they need to choose and use `pwc`
  93. commands effectively. The Skill is generated by your locally installed CLI, so
  94. it matches that CLI version.
  95. ```bash
  96. # Install for this project. Works with Codex, Cursor, OpenCode, and other
  97. # agents that load Skills from .agents/skills.
  98. pwc skills add
  99. # Install for all projects in ~/.agents/skills.
  100. pwc skills add --global
  101. # Also make the Skill available to Claude Code.
  102. pwc skills add --claude
  103. pwc skills add --claude --global
  104. ```
  105. For another coding-agent harness, pass its Skills directory explicitly:
  106. ```bash
  107. pwc skills add --dest /path/to/skills
  108. ```
  109. An existing installation is left untouched. Use `--force` to replace it with
  110. the Skill generated by the currently installed CLI.
  111. ## Quick start
  112. Search the paper catalog:
  113. ```bash
  114. pwc search "retrieval augmented generation"
  115. pwc search "vision transformers" --mode semantic --limit 20
  116. pwc search "attention" --page 2 --limit 10
  117. pwc search "attention" --has-official-implementation
  118. ```
  119. Inspect or read a paper:
  120. ```bash
  121. pwc paper info 1706.03762 # includes tagged organizations
  122. pwc paper info 1706.03762 --include-resources # linked GitHub, project, and HF URLs
  123. pwc paper info 1706.03762 --include-evals # evaluation Markdown table
  124. pwc paper info "Attention Is All You Need"
  125. pwc paper read 1706.03762
  126. ```
  127. Discover papers:
  128. ```bash
  129. pwc paper recent --limit 10
  130. pwc paper recent --implementation-coverage
  131. pwc paper trending --limit 20 --max-age-days 90
  132. pwc paper related 1706.03762 --limit 4
  133. pwc paper lineage list 1706.03762
  134. ```
  135. List and filter papers:
  136. ```bash
  137. pwc paper list --conference NeurIPS --page-size 20
  138. pwc paper list --organization Qwen --order-by date_published
  139. pwc paper list --author "Kaiming He" --search drift --order-by date_published
  140. pwc paper list --task "Image Classification" --method Transformer
  141. pwc paper list --conference "CVPR 2026" --framework PyTorch
  142. pwc paper list --search "diffusion" --start-date 2025-01-01 --end-date 2025-01-31
  143. pwc paper list --order-by citation_count --order-dir desc
  144. pwc paper list --has-official-implementation --implementation-coverage
  145. ```
  146. Explore the research taxonomy and conferences:
  147. ```bash
  148. pwc task list
  149. pwc task list --area Vision
  150. pwc task list --group-by-area # force grouped Markdown when piping/capturing
  151. pwc task --name "scene-text-recognition"
  152. pwc method --name Transformer
  153. pwc method list --area Audio
  154. pwc method list --area General --introduced-year 2017
  155. pwc conference --name "CVPR 2025"
  156. pwc conference list
  157. pwc conference list --year 2025
  158. pwc organization --name NVIDIA
  159. pwc organization list --featured-only
  160. pwc framework --name vLLM
  161. pwc framework list --platform gpu
  162. ```
  163. Browse benchmarks:
  164. ```bash
  165. pwc benchmark list
  166. pwc benchmark list --group-by-area --area Vision
  167. pwc benchmark list --search ImageNet
  168. pwc benchmark list --task OCR
  169. pwc benchmark list --task image-classification --include-descendants
  170. pwc benchmark list --min-eval-count 10 --order-by paper_count --order-dir desc
  171. pwc benchmark --name "SWE-Bench Pro"
  172. pwc benchmark --name wikitext-103 --is-open true --max-parameters 3B
  173. ```
  174. ## Command reference
  175. Every `PAPER` argument accepts a modern or legacy arXiv ID, a numeric
  176. external-paper ID, or a quoted exact paper title. Title matching is
  177. case-insensitive. Ambiguous titles are rejected with the matching IDs instead
  178. of silently selecting one paper.
  179. | Command | Description |
  180. | --- | --- |
  181. | `pwc search QUERY` | Search for papers by title, topic, author, or arXiv ID |
  182. | `pwc paper info PAPER` | Show paper metadata, abstract, predecessors, and successors; optionally include resources or an evaluation Markdown table |
  183. | `pwc paper read PAPER` | Print the stored paper Markdown |
  184. | `pwc paper list` | List and filter papers |
  185. | `pwc paper recent` | List recently published papers |
  186. | `pwc paper trending` | List trending papers |
  187. | `pwc paper related PAPER` | Find related papers |
  188. | `pwc paper lineage list PAPER` | Show linked Markdown sections for a paper's predecessors and successors |
  189. | `pwc task list` | Group tasks by area with paper, benchmark, and evaluation counts; support flat pagination and area filters |
  190. | `pwc task --name NAME` | Show task-page details, including frameworks, sister tasks, common methods, benchmarks, and papers |
  191. | `pwc method --name NAME` | Show details for an exact method name, full name, slug, or ID |
  192. | `pwc method list` | List methods and filter them by research area or introduction year |
  193. | `pwc conference --name NAME` | Show details for an exact conference name, slug, or ID |
  194. | `pwc conference list` | List conferences and filter them by year |
  195. | `pwc organization --name NAME` | Show details for an exact organization name, slug, or ID |
  196. | `pwc organization list` | List organizations, optionally restricting to featured organizations |
  197. | `pwc framework --name NAME` | Show details for an exact framework name, slug, or ID |
  198. | `pwc framework list` | List frameworks by domain, category, or platform |
  199. | `pwc benchmark list` | Group top benchmarks by area and task; support flat filters and task-trend ranking |
  200. | `pwc benchmark --name NAME` | Show a benchmark's top models, parameter counts, papers, scores, and publication dates; optionally apply an inclusive model-size limit |
  201. | `pwc skills add` | Install the version-matched Skill for coding agents |
  202. | `pwc version` | Show the CLI and API contract versions |
  203. Run `pwc --help`, `pwc COMMAND --help`, or
  204. `pwc paper COMMAND --help` for all available options.
  205. ## Output and scripting
  206. Task and benchmark lists use grouped Markdown in an interactive terminal;
  207. `--flat` forces their paginated tables and `--group-by-area` forces Markdown
  208. when piping. Flat benchmark lists use aligned columns in both terminals and
  209. captured output; use `--json` for machine-readable benchmark rows. Other list
  210. and search commands use aligned columns in a terminal and lossless TSV when
  211. captured. Benchmark details use an aligned table in a terminal and Markdown
  212. when piped or captured. Other detail commands use labeled metadata or Markdown.
  213. Paper discovery commands accept `--implementation-coverage` to add official-code
  214. status and the total linked repository count; JSON and `pwc paper info` always
  215. include them. A false official-code status only means the catalog has no linked
  216. official repository. Search and paper-list filters using
  217. `--has-official-implementation` fail closed unless the API confirms the filter.
  218. Add `--json` to any data command for machine-readable output:
  219. ```bash
  220. pwc search "language models" --limit 5 --json
  221. pwc paper info 1706.03762 --include-evals --json
  222. pwc benchmark list --task image-classification --json
  223. ```
  224. JSON responses include a top-level `schema_version` and `data` field:
  225. ```json
  226. {
  227. "schema_version": "v1",
  228. "data": {}
  229. }
  230. ```
  231. The stable exit codes are:
  232. | Code | Meaning |
  233. | ---: | --- |
  234. | `0` | Success |
  235. | `2` | Invalid command or arguments |
  236. | `3` | Network or server failure |
  237. | `4` | Invalid API response |
  238. ## Configuration
  239. `pwc` uses `https://paperswithcode.co/api/v1` by default. To connect to another
  240. compatible v1 API, set `PWC_API_URL`:
  241. ```bash
  242. PWC_API_URL=http://localhost:8000/api/v1 pwc search "transformers"
  243. ```
  244. ## Scope
  245. Public catalog commands are anonymous and read-only. Paper editing is an
  246. explicit authenticated workflow scoped to one confirmed paper; the CLI does
  247. not include general ingestion, image, embedding, CRON, or infrastructure
  248. maintenance commands.
  249. ## Development
  250. Run the CLI from a checkout:
  251. ```bash
  252. uv run --project standalone_cli pwc --help
  253. ```
  254. Run the focused test and contract checks:
  255. ```bash
  256. uv run --project standalone_cli --with pytest pytest standalone_cli/tests
  257. uv run python standalone_cli/scripts/check_contract.py
  258. ```
  259. The implementation lives in [`standalone_cli`](standalone_cli/).
  260. ## Edit a paper with your agent
  261. Requires a server with the paper-edit API enabled. Sign in using the browser
  262. link and confirm the paper and code. This grants one hour of write access to
  263. that paper; each batch publishes immediately under your HF username.
  264. ```bash
  265. pwc auth login --paper 123 # --no-browser for remote machines
  266. pwc paper edit export 123 --output edits.json
  267. # Add explicit operations to edits.json; current is reference material.
  268. pwc paper edit preview 123 --file edits.json
  269. pwc paper edit submit 123 --file edits.json
  270. pwc auth status --paper 123
  271. pwc auth logout --paper 123
  272. ```
  273. For example, add the following to the exported `operations` array, substituting
  274. real existing task/dataset/metric identifiers. A metrics dictionary lets you
  275. submit multiple scores without repeating model and benchmark details:
  276. ```json
  277. {
  278. "section": "evaluations",
  279. "payload": {
  280. "task_id": 1,
  281. "dataset_id": 2,
  282. "model_name": "Example model",
  283. "methodology": "Test split, zero-shot; Table 3",
  284. "metrics": {"Accuracy": 90.5, "F1": 88.1},
  285. "source_url": "https://arxiv.org/abs/2601.00001"
  286. }
  287. }
  288. ```
  289. Add `evaluation_id` alongside `section` to correct a row. Other supported
  290. sections are tasks, methods, repositories, project_pages, hf_artifacts,
  291. x_threads, and source_url. `x_threads` takes `{"x_threads": [post URLs]}` in
  292. display order (at most 10); link only posts in which the paper's authors or
  293. their organization introduce the paper. Section replacements must retain unrelated entries. Evaluation
  294. updates preserve omitted fields. Existing benchmark/metric definitions are
  295. required; evaluation deletion and rank overrides are excluded.
  296. Batches contain at most 50 operations and 1 MiB. Invalid or stale batches write
  297. nothing. Retry uncertain network results with the exact same document and
  298. idempotency key. For a stale revision, export again and reconcile before retrying.
  299. Accounts share a 20-distinct-paper daily limit (UTC) and 30-publication-per-minute
  300. limit across the website and CLI; admins are exempt. `pwc skills add --force`
  301. installs the detailed agent workflow.
  302. Credentials are stored in mode-0600 files inside the mode-0700 directory
  303. `~/.config/pwc/edit-credentials`, keyed by API base URL and paper. Never copy
  304. these files into prompts or repositories. `PWC_API_URL` selects the target API;
  305. editing requires HTTPS, except on loopback development servers, and refuses
  306. redirects. Public read commands never load edit credentials.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.