PinRAG (ndjordjevic/pinrag) is an MCP server listed on the M8ven Trust Index. It scores 71 out of 100, grade C. It declares 6 tools. No publisher has claimed this listing.

Back to MCP Trust IndexM8ven
C
Limited view
71/100

PinRAG

RAG MCP server for PDFs, YouTube, GitHub repos, and Discord exports. Index documents and query with citations via LangChain and Chroma.

→ ndjordjevic/pinrag· listed on glama

Limited view. Automated analysis covers part of this stack. Findings reflect what we verified. Grades reflect the full trust pyramid: code, verification depth, and reputation. New projects cap at C until adoption is earned.

Limited view: static analysis for Python is partially covered.

How we verified

Code Verified⚡ Live Monitored: not connected

Verified is a snapshot. Live keeps it current, and builds your track record.

⚡ Connect GitHub → continuous verification on every pushwhy connect →

Who stands behind it

ndjordjevic

Source: Glama

Is this your MCP?

Claim it to get a verified publisher badge, a free copy of our full audit findings, and direct contact for any high-priority issues we find. Or connect your repo for our deepest verification, Live Monitored: read-only, revoke anytime. What we access →

⚡ Connect GitHub for Live Monitored

Install from

Glamanot scored here

The grade above is for the source repository. Registries can serve a different version, so we mark the ones we were not able to read.

// key findings
✅
No credential exfiltration, no sensitive file access, no obfuscation
Static analysis found nothing flowing your secrets to unexpected places.
✅
Open source with a license and README
Anyone can audit the code, the license is declared, and the publisher documents what it does.
🔐
You'll be asked for 5 credentials: ANTHROPIC_API_KEY, OPENROUTER_API_KEY, OPENAI_API_KEY, GITHUB_TOKEN, CEREBRAS_API_KEY
These are read from process.env at runtime. Make sure you trust where they’ll be sent.
// environment variables
To run this server yourself, you supply these values. They go in your own MCP client configuration and stay on your machine. The secret label means the value is sensitive, not that the server mishandles it.
configTQDM_DISABLE
configPINRAG_LLM_PROVIDERopenrouter openrouter, openai, anthropic, or cerebras
configOPENAI_BASE_URL(optional) Override OpenAI API base URL (e.g. https://openrouter.ai/api/v1 with OPENAI_API_KEY set to your OpenRouter key for vision or other OpenAI-compatible calls).
🔐 secretANTHROPIC_API_KEY(required for Anthropic LLM or Anthropic YouTube vision) Required when PINRAG_LLM_PROVIDER=anthropic, PINRAG_EVALUATOR_PROVIDER=anthropic, or YouTube vision with PINRAG_YT_VISION_PROVIDER=anthropic.
🔐 secretOPENROUTER_API_KEY"": "your-openrouter-api-key-here",
🔐 secretOPENAI_API_KEY(required for OpenAI LLM or OpenAI YouTube vision) Required when PINRAG_LLM_PROVIDER=openai, or when PINRAG_YT_VISION_ENABLED=true and PINRAG_YT_VISION_PROVIDER=openai.
configPINRAG_USE_RERANKfalse Set to true to enable FlashRank re-ranking: fetch more chunks, re-score locally, pass top N to the LLM. Requires pip install pinrag[rerank] and no API key.
configPINRAG_RETRIEVE_K20 Retrieval pool size when re-ranking is off. When re-ranking is on, PINRAG_RERANK_RETRIEVE_K falls back to this value if unset, then results are cut to PINRAG_RERANK_TOP_N.
configPINRAG_CEREBRAS_BASE_URL
configPINRAG_LLM_MODEL
configOPENROUTER_APP_TITLEPinRAG App title (X-Title). Override to label usage in the OpenRouter dashboard. PinRAG copies this into OPENROUTER_X_OPEN_ROUTER_TITLE for the SDK.
configOPENROUTER_HTTP_REFERER
configOPENROUTER_X_OPEN_ROUTER_TITLEOPENROUTER_APP_TITLE PinRAG App title (X-Title). Override to label usage in the OpenRouter dashboard. PinRAG copies this into for the SDK.
configPINRAG_OPENROUTER_MODEL_FALLBACKS
configPINRAG_LLM_MODEL_FALLBACKS
configPINRAG_OPENROUTER_SORT
configPINRAG_OPENROUTER_PROVIDER_ORDER
configPINRAG_EVALUATOR_PROVIDERANTHROPIC_API_KEY (required for Anthropic LLM or Anthropic YouTube vision) Required when PINRAG_LLM_PROVIDER=anthropic, =anthropic, or YouTube vision with PINRAG_YT_VISION_PROVIDER=anthropic.
configPINRAG_EMBEDDING_MODELLocal embeddings — Nomic (, default nomic-embed-text-v1.5); no API key; first run downloads model weights (~270 MB, cached)
configPINRAG_PERSIST_DIR"": "/absolute/path/to/your/pinrag-data"
configPINRAG_COLLECTION_NAMEMCP tools — add_document_tool, query_tool, list_documents_tool, remove_document_tool, set_document_tag_tool, list_collections_tool; optional collection on tools overrides for that call
configPINRAG_CHUNK_SIZE1000 Text chunk size (chars)
configPINRAG_CHUNK_OVERLAP200 Chunk overlap (chars)
configPINRAG_RERANK_RETRIEVE_KPINRAG_RETRIEVE_K 20 Retrieval pool size when re-ranking is off. When re-ranking is on, falls back to this value if unset, then results are cut to PINRAG_RERANK_TOP_N.
configPINRAG_RERANK_TOP_NPINRAG_RETRIEVE_K 20 Retrieval pool size when re-ranking is off. When re-ranking is on, PINRAG_RERANK_RETRIEVE_K falls back to this value if unset, then results are cut to .
configPINRAG_USE_MULTI_QUERYfalse Set to true to generate alternative phrasings of the user query via LLM, retrieve per variant, and merge (unique union). Improves recall for terse or ambiguous queries.
configPINRAG_MULTI_QUERY_COUNT4 Number of alternative queries to generate (default 4, max 10). The original query is still included in retrieval when merging.
configPINRAG_USE_PARENT_CHILDfalse Set to true to embed small chunks and return larger parent chunks (supported for PDF, GitHub, YouTube, and Discord indexing—not plain .txt). Requires re-indexing.
configPINRAG_PARENT_CHUNK_SIZE2000 Parent chunk size (chars) when PINRAG_USE_PARENT_CHILD=true.
configPINRAG_CHILD_CHUNK_SIZE800 Child chunk size (chars) when PINRAG_USE_PARENT_CHILD=true.
configPINRAG_RESPONSE_STYLEQuery with filters query_tool — required query. Optional document_id, tag, document_type, page_min / page_max (PDF ranges), response_style (thorough or concise; leave empty to use ).
configPINRAG_STRUCTURE_AWARE_CHUNKINGtrue Apply structure-aware chunking heuristics for code/table boundaries
configPINRAG_VERBOSE_LOGGINGFor MCP-side introspection, set =true to surface detailed phase events in notifications/message (e.g., YouTube transcript load, whether vision runs, and chunk upsert milestones).
🔐 secretGITHUB_TOKEN(optional) Personal access token for GitHub API. Required for private repos; increases rate limits for public repos.
configPINRAG_GITHUB_MAX_FILE_BYTES524288 (512 KB) Skip files larger than this when indexing GitHub repos.
configPINRAG_GITHUB_DEFAULT_BRANCHmain Default branch when not specified in the GitHub URL.
configPINRAG_WEB_USER_AGENTPinRAGBot/<version> (+https://github.com/ndjordjevic/pinrag) HTTP User-Agent header for web indexing. Some sites block generic bots; override if needed.
configPINRAG_PLAINTEXT_MAX_FILE_BYTES524288 (512 KB) Skip plain .txt files larger than this when indexing.
configPINRAG_YT_PROXY_HTTP_URL(none) HTTP proxy URL for transcript fetches (e.g. http://user:pass@proxy:80). Use when YouTube blocks your IP.
configPINRAG_YT_PROXY_HTTPS_URL(none) HTTPS proxy URL for transcript fetches. Same as HTTP when using a generic proxy.
configPINRAG_YT_VISION_ENABLEDOPENAI_API_KEY (required for OpenAI LLM or OpenAI YouTube vision) Required when PINRAG_LLM_PROVIDER=openai, or when =true and PINRAG_YT_VISION_PROVIDER=openai.
configPINRAG_YT_VISION_PROVIDEROPENAI_API_KEY (required for OpenAI LLM or OpenAI YouTube vision) Required when PINRAG_LLM_PROVIDER=openai, or when PINRAG_YT_VISION_ENABLED=true and =openai.
configPINRAG_YT_VISION_MODEL(per provider) If unset: OpenAI gpt-4o-mini, Anthropic claude-sonnet-4-6, OpenRouter google/gemini-2.5-flash. Use a vision-capable id. Legacy alias: PINRAG_VISION_MODEL.
configPINRAG_VISION_MODELPINRAG_YT_VISION_MODEL (per provider) If unset: OpenAI gpt-4o-mini, Anthropic claude-sonnet-4-6, OpenRouter google/gemini-2.5-flash. Use a vision-capable id. Legacy alias: .
configPINRAG_YT_VISION_MAX_FRAMES8 Download + frame path only (openai / anthropic): cap analyzed keyframes after scene detect. Ignored for openrouter (single video request).
configPINRAG_YT_VISION_MIN_SCENE_SCORE27.0 Download + frame path only: PySceneDetect AdaptiveDetector threshold (higher → fewer cuts). Ignored for openrouter.
configPINRAG_YT_VISION_IMAGE_DETAILlow OpenAI frame path only: low, high, or auto for image_url.detail. Ignored for anthropic (full frames) and openrouter (video_url).
🔐 secretCEREBRAS_API_KEY(required for Cerebras LLM) Required when PINRAG_LLM_PROVIDER=cerebras. Get a key from the [Cerebras cloud console](https://cloud.cerebras.ai).
Deployment configuration, supplied by whoever hosts the server. Users are not asked for these.
deployOPENROUTER_APP_URL
// quality suggestions

Tool annotations

No tools have read-only/destructive annotations

Add readOnlyHint or destructiveHint annotations to every tool so hosts can warn users before invoking.

All four hints declared on every tool

6/6 tools missing one or more hints — query_tool (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); add_document_tool (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint); list_documents_tool (missing: readOnlyHint, destructiveHint, idempotentHint, openWorldHint), +3 more. OpenAI's directory rejects tools where any of the four hints are missing or non-boolean.

For every tool, set all four hints (readOnlyHint, destructiveHint, idempotentHint, openWorldHint) to explicit true/false values that match the handler’s actual behaviour.

Tool handlers catch errors

5/6 tool handlers wrap calls in try/catch (83%)

Wrap each tool handler body in try/catch and return a structured error response.

Tests exist

No test files found

Add tests that exercise each declared tool.

Claim the listing to review these findings one by one and send us a correction where you disagree, straight to the team. Claiming also means we tell you when the grade moves, and reach you first if we find anything urgent.

// full audit trail
The findings above are the summary. The full trail, every check we ran, each deduction, the network hosts observed and the dependency advisories, goes to verified publishers, along with an alert whenever a new one lands. Verified publishers can also review each finding and dispute it in one click. Publisher corrections have sharpened several of our checks this month, because the maintainer knows the codebase better than any scanner.
// improvement guidance — verified publishers only
We have 4 concrete improvements we can share with the publisher of this MCP. Each comes with specific guidance to raise the trust score.
// embed badge in your README
[![M8ven Score](https://m8ven.ai/badge/mcp/ndjordjevic/pinrag)](https://m8ven.ai/mcp/ndjordjevic/pinrag)
Shows your grade and updates automatically. Prefer no grade? Append ?variant=verified to the badge URL.
commit: 3f3336e69bbe45e193a96210de5675568a383c1d
code hash: 885ef7821c3d0a235238dd5c67197bba85eee738428cfa1dde074da679e727ee
view raw JSON →
Check MCPs from inside your assistant
Tool Check · MCP

Vetting this one by hand? Tool Check is an MCP that scores other MCPs. Add it once and ask Claude, ChatGPT, or any MCP client to grade a server, surface CVEs, check the publisher, and suggest safer alternatives — before you install.

https://m8ven.ai/api/mcp/tool-check
check_toolsearch_toolscompare_toolsrecommend_alternativescheck_publisherreport_concern
How to add it →Free · no account needed · works in any MCP client