About litra-paper-search
litra-paper-search was an MCP server for searching academic papers through Litra.ai, aimed at research workflows. It offered relevance scoring, summaries, author lookup and credit checks, but is no longer maintained and cannot work because the public API was discontinued.
Key features
- 01Natural-language paper search
- 02AI relevance scores and summaries
- 03Author profiles and publication search
- 04Credit balance and plan checks
Use cases
- 01Find papers for a literature review
- 02Look up an author's publications
- 03Check Litra.ai account credits
MCP support
It exposed paper search, author search and credit-check tools to MCP clients. The server no longer functions because Litra.ai discontinued the public API.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
npx -y litra-paper-search-mcpLITRA_API_KEYSecretLitra.ai API key (get it at litra-ai.com/account)Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
3Saved tool list · updated 2w ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| query | string | Yes | Natural language search query |
| year_to | integer | No | Maximum publication year (inclusive). NOTE: Year filtering is currently experimental. See year_from for details on the known limitation (Litra.ai issue #1068). |
| year_from | integer | No | Minimum publication year (inclusive). NOTE: Year filtering is currently experimental. The pilot search stage does not yet 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. |
| max_results | integer | No | Maximum number of papers (10-30, default: 30) |
Details
From the project READMElitra-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⚠️ This project is no longer maintained.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.
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
- Sign in at litra-ai.com/account
- Navigate to the API Keys section
- 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.
| Parameter | Type | Required | Description |
|---|---|---|---|
query | string | Yes | Natural language search query (1-500 chars) |
max_results | number | No | Maximum number of papers (10-30, default: 30) |
year_from | number | No | Minimum publication year (inclusive). Experimental – see Known limitations. |
year_to | number | No | Maximum publication year (inclusive). Experimental – see Known limitations. |
Example usage:
Search for recent papers on transformer architectures in NLP from 2023 onwards
Known limitations (year filtering): The
year_from/year_tofilters 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 theiryearfield.
search_author
Search academic papers by author name.
| Parameter | Type | Required | Description |
|---|---|---|---|
name | string | Yes | Author name |
limit | number | No | Maximum number of papers (1-50, default: 20) |
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.
| Plan | Rate Limit |
|---|---|
| Mini | 10 requests/min |
| Standard | 30 requests/min |
| Pro | 60 requests/min |
| Max | 120 requests/min |
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.git
cd litra-paper-search-mcp
npm install
npm 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
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.