{"_id":"citeproof-mcp","_rev":"4-c7e109bd470d969431b374fc040a5113","name":"citeproof-mcp","dist-tags":{"latest":"0.3.0"},"versions":{"0.0.0-stage":{"name":"citeproof-mcp","version":"0.0.0-stage","_id":"citeproof-mcp@0.0.0-stage","stub":true,"maintainers":[{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"}],"dist":{"shasum":"2414c83244a5229e7c9941fa96e68de06d8d2494","tarball":"https://registry.npmjs.org/citeproof-mcp/-/citeproof-mcp-0.0.0-stage.tgz","fileCount":2,"integrity":"sha512-uimZ/cKfIWmk4O7gTP5Bzh6sW2SZoR3ORlRF8rf7DLkfumNW41nLz5R3tOx+uMqKYOLvPgjLr7uRYgTEhSMJog==","signatures":[{"sig":"MEYCIQDUWhkbKyCWts8i/pyPQWtsO2Tsg3a2iyG5OjfP+srj7wIhAKfcXp8y6+WOsQaHmvZZjOQT9d5xsSE7v5zdbdVbebMc","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":335},"_npmUser":{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"},"description":"Temporary package placeholder for staged publishing","directories":{}},"0.2.0":{"name":"citeproof-mcp","version":"0.2.0","keywords":["mcp","model-context-protocol","claude","systematic-review","prisma","citations"],"author":{"name":"Aryan Pardeshi"},"license":"MIT","_id":"citeproof-mcp@0.2.0","maintainers":[{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"}],"homepage":"https://citeproof.top/mcp","bugs":{"url":"https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/issues"},"bin":{"citeproof-mcp":"bin/citeproof-mcp.js"},"dist":{"shasum":"69ba89682258c06e7ab65dfd457a885205368114","tarball":"https://registry.npmjs.org/citeproof-mcp/-/citeproof-mcp-0.2.0.tgz","fileCount":9,"integrity":"sha512-SqNIfIE4d9tbhRozl3EkAVxZxy47JxecLMSQpwuqkF2s+js1gSOvAx5rzjKI+HSSX/SHbq72suOkqCFbAwgTzg==","signatures":[{"sig":"MEYCIQDaawZyQnJ+pPjpAd+Yp67mEjc1gjAGtU0GJj89e7qLnQIhALQKSTEwmUkWUgDPxnQ2IoVPcqkWZ4XOp845MzdKPT0H","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"},{"sig":"MEYCIQDNppgMloGDf2URel5k9mrm4IxK/bZ6QN1lFJmPMDTV9AIhAKHs8dv5P+NTNor4qxIdKEeN5xFvrlhHjhSHWgMPkwcv","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":29186},"engines":{"node":">=18"},"scripts":{"test":"node --test"},"_npmUser":{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"},"repository":{"url":"git+https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP.git","type":"git","directory":"MCP/npm"},"_npmVersion":"11.14.1","description":"Audit layer for AI literature reviews: run the CiteProof MCP server over stdio with npx. Records PRISMA counts, screening decisions, evidence checks and citation checks supplied by your research agent.","directories":{},"_nodeVersion":"24.19.0","_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/citeproof-mcp_0.2.0_1792879102840_0.9813142910679793","host":"s3://npm-registry-packages-npm-production"}},"0.2.1":{"name":"citeproof-mcp","version":"0.2.1","keywords":["mcp","model-context-protocol","claude","systematic-review","prisma","citations"],"author":{"name":"Aryan Pardeshi"},"license":"MIT","_id":"citeproof-mcp@0.2.1","maintainers":[{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"}],"homepage":"https://citeproof.top/mcp","bugs":{"url":"https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/issues"},"bin":{"citeproof-mcp":"bin/citeproof-mcp.js"},"dist":{"shasum":"32f0a78c963181d0e26e4d81d530cd020842ce59","tarball":"https://registry.npmjs.org/citeproof-mcp/-/citeproof-mcp-0.2.1.tgz","fileCount":34,"integrity":"sha512-Cjk8rZUM4ubg/olHSmhrZ3YpReZ+uzsUnrkUk+lum/QwkzWgLfBcbRQ5u42BVctCJ6gH1CahujoxlGDp0GjYMw==","signatures":[{"sig":"MEUCIQCl3oNdrIm7Qw3Z4U450FfnTv3WM4zceXs5AJKvP9nxbwIgKWj0j15ctdmUhfY1pAzT18uIFASjxNJG+FsXahpfis8=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"},{"sig":"MEUCIHNYSyT3z+AQYoFw0G7eyQFgM/8tnNM2zna6yRYWrU8jAiEAxPMbsyndOSJig90oxKB+u7TjKHIyBms+noqSSudCtIA=","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":329608},"engines":{"node":">=18"},"scripts":{"test":"node --test","prepack":"node scripts/bundle-python.js"},"_npmUser":{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"},"repository":{"url":"git+https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP.git","type":"git","directory":"MCP/npm"},"_npmVersion":"11.14.1","description":"Audit layer for AI literature reviews: run the CiteProof MCP server over stdio with npx. Records PRISMA counts, screening decisions, evidence checks and citation checks supplied by your research agent.","directories":{},"_nodeVersion":"24.19.0","_hasShrinkwrap":false,"_npmOperationalInternal":{"tmp":"tmp/citeproof-mcp_0.2.1_1792883962154_0.34766277468840046","host":"s3://npm-registry-packages-npm-production"}},"0.3.0":{"_id":"citeproof-mcp@0.3.0","bin":{"citeproof-mcp":"bin/citeproof-mcp.js"},"bugs":{"url":"https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/issues"},"dist":{"shasum":"0c107044984f2f7fed3ec301748a15d1702231ba","tarball":"https://registry.npmjs.org/citeproof-mcp/-/citeproof-mcp-0.3.0.tgz","fileCount":38,"integrity":"sha512-2ccjE0HQlCADzFqephJ/VMex71vUyu2MjKUVxAWlkl487kbGfs8J0INQ22ConJP6ieFlX+Hb+idWdoPD+lzDIg==","signatures":[{"sig":"MEQCIALuicbkqxvETOOGu2cC6tqccpzwF+fUDWo/PFghpdF3AiAI5G+hWSC7CVT2RugNEQPA7VXvuty1VSrtTrsbdyIrvg==","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"},{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEUCIQDHOF/Xl/N91pffOvSkRWid3iqBhD46Y6D4wXEMduolwQIgVEwxIeY5m13Az5i3d+hZ/ij54nDarR8qNXOa86PLCrs="}],"unpackedSize":513406},"name":"citeproof-mcp","author":{"name":"Aryan Pardeshi"},"engines":{"node":">=18"},"gitHead":"96fe8da599acb50b8c174b70cd632e1e4645867c","license":"MIT","scripts":{"test":"node --test","prepack":"node scripts/bundle-python.js"},"version":"0.3.0","_npmUser":{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"},"homepage":"https://citeproof.top/mcp","keywords":["mcp","model-context-protocol","claude","systematic-review","prisma","citations"],"repository":{"url":"git+https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP.git","type":"git","directory":"MCP/npm"},"_npmVersion":"11.14.1","description":"Audit layer for AI literature reviews: run the CiteProof MCP server over stdio with npx. Records PRISMA counts, screening decisions, evidence checks and citation checks supplied by your research agent.","directories":{},"maintainers":[{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"}],"_nodeVersion":"24.19.0","_hasShrinkwrap":false,"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/citeproof-mcp_0.3.0_1793361970707_0.7945961859610058"}}},"time":{"created":"2026-10-24T21:57:27.445Z","modified":"2026-10-30T12:06:11.051Z","0.0.0-stage":"2026-10-24T21:57:27.445Z","0.2.0":"2026-10-24T21:58:22.927Z","0.2.1":"2026-10-24T23:19:22.254Z","0.3.0":"2026-10-30T12:06:10.875Z"},"bugs":{"url":"https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/issues"},"author":{"name":"Aryan Pardeshi"},"license":"MIT","homepage":"https://citeproof.top/mcp","keywords":["mcp","model-context-protocol","claude","systematic-review","prisma","citations"],"repository":{"url":"git+https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP.git","type":"git","directory":"MCP/npm"},"description":"Audit layer for AI literature reviews: run the CiteProof MCP server over stdio with npx. Records PRISMA counts, screening decisions, evidence checks and citation checks supplied by your research agent.","maintainers":[{"name":"aryanpardeshi","email":"aryanpardeshi605@gmail.com"}],"readme":"# citeproof-mcp (npm launcher)\r\n\r\n**Audit layer for AI literature reviews.** This package is a thin Node launcher for the [CiteProof MCP](https://citeproof.top/mcp) server. It ships the Python server source and runs it with a Python 3.12+ that is already on your machine, or with `uvx` if that is available (installing [uv](https://docs.astral.sh/uv/) automatically if it is missing and no Python works). Any MCP client can add CiteProof the same way as any other MCP server.\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"citeproof\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"citeproof-mcp\"]\r\n    }\r\n  }\r\n}\r\n```\r\n\r\nCiteProof MCP records and checks what your research agent supplies: review protocols, PRISMA counts, screening decisions, extracted evidence, risk-of-bias records, DOI and quote checks, and a reproducible export bundle. **It does not search the literature or read papers**, and it contains no language model. Your agent does the research; CiteProof keeps it auditable. No API key is required.\r\n\r\nCiteProof runs **entirely on your computer**. There is no hosted server, no account, and no upload of review data.\r\n\r\n## Requirements\r\n\r\n- Node.js 18 or newer.\r\n- Either Python 3.12+ with the `mcp`, `pydantic` and `httpx` packages (the launcher probes `python3`, `python` and, on Windows, `py -3` on your `PATH` and runs the bundled server without further downloads), or network access on the first run so the launcher can download uv, Python 3.12 and the server wheel. Either way, `npx` needs the npm registry the first time unless the package is already cached.\r\n- macOS, Linux or Windows.\r\n\r\n## Install\r\n\r\n### Claude Science\r\n\r\n1. Open **Customize → Connectors**.\r\n2. Choose **Add connector → Local command**.\r\n3. Set **Name** to `citeproof` and the **Command** to:\r\n\r\n   ```bash\r\n   npx -y citeproof-mcp\r\n   ```\r\n\r\n4. Save and start a new conversation.\r\n\r\nThe npm package contains the server source, so inside Claude Science's sandbox the launcher runs it with the sandbox's own `python3` (which already has `mcp`, `pydantic` and `httpx`). Claude Science stages the npm package for you, and when `uvx` is not available the launcher needs no uv, no download and no access to astral.sh, GitHub or PyPI. If the sandbox has `uvx`, the launcher prefers it; set `CITEPROOF_MCP_PYTHON=python3` to select the sandbox's Python. If the sandbox has no usable Python, see [Troubleshooting](#troubleshooting).\r\n\r\n### Claude Desktop\r\n\r\nAdd this to `claude_desktop_config.json` (Settings → Developer → Edit Config):\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"citeproof\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"citeproof-mcp\"]\r\n    }\r\n  }\r\n}\r\n```\r\n\r\nRestart Claude Desktop after saving.\r\n\r\n### Claude Desktop extension\r\n\r\nIf you prefer a one-click install, download `citeproof.mcpb` from the [latest GitHub release](https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/releases/latest) and open it with Claude Desktop. The bundle contains the same launcher and runs with Claude Desktop's built-in Node. The launcher uses the bundled source with a compatible local Python when available; otherwise its uvx fallback may need network access to download the server wheel.\r\n\r\n### Claude Code\r\n\r\n```bash\r\nclaude mcp add citeproof -- npx -y citeproof-mcp\r\n```\r\n\r\n### Cursor, VS Code, Windsurf\r\n\r\nUse the same generic `mcpServers` block shown at the top of this page in the client's MCP configuration file.\r\n\r\n- **Cursor:** `~/.cursor/mcp.json` (or `.cursor/mcp.json` in a project).\r\n- **VS Code:** the `mcp.json` used by your MCP extension; the built-in support uses `\"servers\"` instead of `\"mcpServers\"` — check your client's documentation.\r\n- **Windsurf:** `~/.codeium/windsurf/mcp_config.json`.\r\n\r\n> These three clients are **not yet tested** with this launcher. The command itself is plain stdio and should work anywhere `npx` works, but please open an issue if a client needs a different shape.\r\n\r\n### claude.ai web\r\n\r\nThe claude.ai web custom connector UI expects a remote MCP server reachable over HTTP. CiteProof is a local stdio server and **is not supported there**. Use Claude Desktop, Claude Science, Claude Code or another desktop MCP client.\r\n\r\n## Privacy\r\n\r\n- The server and the launcher run locally. Review records are written to a local SQLite database under `CITEPROOF_MCP_HOME` (default `~/.citeproof-mcp`) and never leave your computer.\r\n- The only network calls the server makes are DOI lookups in `verify_dois`: Crossref first, then doi.org, to check that a DOI is registered and not retracted. The DOIs you ask it to check are sent to those services; nothing else is.\r\n- The launcher runs the server locally. It uses a Python that is already on your machine when it can, and only downloads uv, Python and the server wheel when neither `uvx` nor a suitable Python is available. It has no telemetry and no analytics.\r\n\r\n## What the launcher does\r\n\r\n1. If `CITEPROOF_MCP_PYTHON` is set (or `--python <path>` is passed), it runs the bundled server with that interpreter. The interpreter must be Python 3.12+ with `mcp`, `pydantic` and `httpx`.\r\n2. Otherwise it looks for `uvx` in this order: `CITEPROOF_UVX`, your `PATH`, `<CITEPROOF_UV_HOME>/bin`, then the usual uv install locations (`~/.local/bin/uvx`, `~/.cargo/bin/uvx`, and the Windows equivalents under `%USERPROFILE%` and `%LOCALAPPDATA%`). If found, it starts `uvx --from <source> --python 3.12 citeproof-mcp`.\r\n3. If `uvx` is missing, it probes `python3`, then `python` (and `py -3` on Windows) with a 15-second timeout, checking that the interpreter is Python 3.12+ and can import `mcp`, `pydantic` and `httpx`. The first one that passes runs the bundled server: `<python> -m citeproof_mcp`, with the packaged `python/` folder prepended to `PYTHONPATH` and `PYTHONUNBUFFERED=1`. No download and no network are needed.\r\n4. If no interpreter passes, it runs Astral's official installer non-interactively (`https://astral.sh/uv/install.sh`, or `install.ps1` on Windows), prints progress to **stderr**, and then starts uvx. It does not edit your shell profiles. The default source is this repository at the tag matching the launcher version, so uv also downloads Python 3.12 if your machine does not have it.\r\n5. It forwards SIGINT/SIGTERM to the server and exits with the server's exit code. Extra arguments after the launcher flags are passed through to the server unchanged.\r\n\r\nstdout is reserved for the MCP protocol. The launcher never writes to it during a session; all diagnostics go to stderr. There is no telemetry.\r\n\r\n## Environment variables\r\n\r\n| Variable | Purpose |\r\n| --- | --- |\r\n| `CITEPROOF_MCP_HOME` | Data directory for the server's local SQLite records (default `~/.citeproof-mcp`). |\r\n| `CITEPROOF_MCP_PYTHON` | Python 3.12+ interpreter to use with the bundled server. Skips uv entirely. The `--python <path>` flag does the same and wins over the variable. |\r\n| `CITEPROOF_MCP_SOURCE` | Override the `uvx --from` value, for example `citeproof-mcp==0.3.0` once the server is on PyPI, or an absolute path to a local checkout of `MCP/`. |\r\n| `CITEPROOF_UV_HOME` | Keep everything uv writes in one folder: `cache/`, `tools/`, `python/` and `bin/`. Useful inside a sandbox. The `--uv-home <dir>` flag does the same and wins over the variable. |\r\n| `CITEPROOF_NO_AUTO_INSTALL` | Set to `1` to never install uv automatically. The launcher then exits with a manual install link. |\r\n| `CITEPROOF_UVX` | Full path to the `uvx` executable, if it is not on `PATH`. |\r\n\r\nExample with a pinned source and a custom data directory:\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"citeproof\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"citeproof-mcp\"],\r\n      \"env\": {\r\n        \"CITEPROOF_MCP_HOME\": \"/Users/you/Documents/citeproof\",\r\n        \"CITEPROOF_MCP_SOURCE\": \"citeproof-mcp==0.3.0\"\r\n      }\r\n    }\r\n  }\r\n}\r\n```\r\n\r\n## Troubleshooting\r\n\r\n**`uv (uvx) was not found ... No Python 3.12+ with the mcp, pydantic and httpx packages was found either.`** The launcher could not find uvx and could not find a Python with the three packages. Either install uv (below), or set `CITEPROOF_MCP_PYTHON` to a Python 3.12+ that has `mcp`, `pydantic` and `httpx`:\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"citeproof\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"citeproof-mcp\"],\r\n      \"env\": { \"CITEPROOF_MCP_PYTHON\": \"/usr/bin/python3\" }\r\n    }\r\n  }\r\n}\r\n```\r\n\r\n`--python /usr/bin/python3` does the same. The launcher then runs the server from the package's own `python/` folder; it needs no uv and downloads nothing. Nothing is printed to stdout.\r\n\r\n**Claude Science says a file is not readable inside the sandbox.** This only applies when the launcher falls back to uv: uv normally writes to `~/.local/share/uv`, `~/.cache/uv` and similar folders, which may sit outside the connector's workspace. Set `CITEPROOF_UV_HOME` to a folder inside the connector's workspace, for example:\r\n\r\n```json\r\n{\r\n  \"mcpServers\": {\r\n    \"citeproof\": {\r\n      \"command\": \"npx\",\r\n      \"args\": [\"-y\", \"citeproof-mcp\", \"--uv-home\", \"/path/to/workspace/.citeproof-uv\"]\r\n    }\r\n  }\r\n}\r\n```\r\n\r\nThe launcher creates `cache/`, `tools/`, `python/` and `bin/` there and points `UV_CACHE_DIR`, `UV_TOOL_DIR`, `UV_PYTHON_INSTALL_DIR` and (for the auto-installer) `UV_INSTALL_DIR` at them.\r\n\r\n**uv is missing and auto-install is disabled.** Install uv manually, then rerun the client:\r\n\r\n```bash\r\n# macOS / Linux\r\ncurl -LsSf https://astral.sh/uv/install.sh | sh\r\n\r\n# Windows (PowerShell)\r\npowershell -ExecutionPolicy Bypass -c \"irm https://astral.sh/uv/install.ps1 | iex\"\r\n```\r\n\r\nFull instructions: <https://docs.astral.sh/uv/getting-started/installation/>. You can also point `CITEPROOF_UVX` at the installed binary.\r\n\r\n## Links\r\n\r\n- Server, tools and documentation: <https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/tree/main/MCP>\r\n- Claude Desktop extension (`citeproof.mcpb`): <https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/releases/latest>\r\n- Issues: <https://github.com/Aryan-Pardeshi/CiteProof-AI-MCP/issues>\r\n\r\n## Licence\r\n\r\nMIT. See [LICENSE](LICENSE).\r\n","readmeFilename":"README.md"}