Litra Paper Search

litra-paper-search-mcp

⚠️ This project is no longer maintained.

Litra AI discontinued its public API and MCP server on 2026-09-04. The API endpoints this server calls no longer exist, so it cannot work even with a valid API key.

Why: what makes Litra AI useful is the way it shows papers — citation trees, lineage of references, year distributions, research maps. MCP can only carry text, so none of that reaches you. A plain list of papers is something other services already provide. We chose to focus on the web app, where the value actually lands.

Please use the web app at https://litra-ai.com instead.

This repository stays online, read-only, for reference.

[npm version] [License: MIT]

MCP server for Litra.ai – AI-powered academic paper search with relevance scoring and summarization.

Features

• search_papers – Search academic papers with AI-powered relevance scoring (1-10) and summarization. Year filtering is available but currently experimental (see Known limitations).
• search_author – Search papers by author name. Returns author profile (affiliation, h-index) and publications.
• check_credits – Check your Litra.ai credit balance and subscription plan.

Prerequisites

• Node.js 18 or higher
• A Litra.ai account with a paid plan (Mini or above)

Get Your API Key

1. Sign in at litra-ai.com/account
2. Navigate to the API Keys section
3. Click Create API Key and copy the key (it is shown only once)

Setup

Claude Code

claude mcp add litra-paper-search -- npx -y litra-paper-search-mcp

Then set the environment variable LITRA_API_KEY in your shell or Claude Code config.

Claude Desktop

Add to your claude_desktop_config.json:

{  "mcpServers": {    "litra-paper-search": {      "command": "npx",      "args": ["-y", "litra-paper-search-mcp"],      "env": {        "LITRA_API_KEY": "litra_xxxxxxxxxxxx"      }    }  }}

Cursor

Add to .cursor/mcp.json:

{  "mcpServers": {    "litra-paper-search": {      "command": "npx",      "args": ["-y", "litra-paper-search-mcp"],      "env": {        "LITRA_API_KEY": "litra_xxxxxxxxxxxx"      }    }  }}

Windsurf

Add to .windsurf/mcp.json:

{  "mcpServers": {    "litra-paper-search": {      "command": "npx",      "args": ["-y", "litra-paper-search-mcp"],      "env": {        "LITRA_API_KEY": "litra_xxxxxxxxxxxx"      }    }  }}

Cline

Add to .cline/mcp_settings.json:

{  "mcpServers": {    "litra-paper-search": {      "command": "npx",      "args": ["-y", "litra-paper-search-mcp"],      "env": {        "LITRA_API_KEY": "litra_xxxxxxxxxxxx"      }    }  }}

GitHub Copilot (VS Code)

Add to .vscode/mcp.json:

{  "servers": {    "litra-paper-search": {      "type": "stdio",      "command": "npx",      "args": ["-y", "litra-paper-search-mcp"],      "env": {        "LITRA_API_KEY": "litra_xxxxxxxxxxxx"      }    }  }}

Tools

search_papers

Search academic papers with AI-powered relevance scoring and summarization.

Example usage:

Search for recent papers on transformer architectures in NLP from 2023 onwards

Known limitations (year filtering): The year_from / year_to filters are still experimental. The pilot search stage currently does not apply the year filter, so some papers outside the specified range may still appear in results. Full support is tracked in Litra.ai issue #1068. As a workaround, you may need to post-filter the returned papers by their year field.

search_author

Search academic papers by author name.

Example usage:

Find papers by Yoshua Bengio

check_credits

Check your Litra.ai credit balance and subscription plan. No parameters required.

Pricing & Rate Limits

API access requires a paid Litra.ai subscription (Mini or above). Each search consumes credits from your existing subscription pool.

See litra-ai.com/pricing for plan details and pricing.

Response Language

All API responses — including paper titles, summaries, and research overviews — are returned in English only.

This is an intentional design choice for the MCP server and public API:

• LLM accuracy: Large language models perform reasoning, analysis, and tool use more reliably in English. Returning structured data in English ensures the highest quality when AI agents process search results for downstream tasks (e.g., literature reviews, citation analysis, research synthesis).
• Token efficiency: English responses consume fewer tokens than translated equivalents, reducing both latency and cost for LLM-based workflows.
• No multilingual plans for the API: This MCP server is designed specifically for AI agent consumption. There are no plans to add other response languages. If you need Japanese or other languages, use the Litra.ai web interface directly, which provides full Japanese support including translated titles and summaries.

If your AI workflow requires output in a specific language, the AI client can translate the English response as a post-processing step — this is more reliable than asking the API to translate, since the LLM can adapt the translation to your specific context.

Troubleshooting

"LITRA_API_KEY environment variable is required"

Set the LITRA_API_KEY environment variable in your MCP client configuration. See the Setup section for examples.

"Invalid API key"

Your API key may be incorrect or revoked. Generate a new key at litra-ai.com/account.

"Insufficient credits"

Your credit balance is too low. Top up at litra-ai.com/account.

"Paid subscription required"

API access requires a paid plan (Mini or above). Subscribe at litra-ai.com/pricing.

"Rate limit exceeded"

You've exceeded your plan's rate limit. Wait a moment and try again, or upgrade your plan for a higher limit.

Development

git clone https://github.com/akkylab/litra-paper-search-mcp.gitcd litra-paper-search-mcpnpm installnpm run build

To test locally with Claude Code:

claude mcp add litra-paper-search -- node /path/to/litra-paper-search-mcp/dist/index.js

License

MIT