Ragstack

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/hatmanstack-ragstack

About

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 25, 2026.

Tools

9 read · 4 write · 0 delete · 2 not yet classified

Secrets

1 (API keys, tokens or passwords)

Install scripts

None found

Runs as

Python package

Tools (15)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

• analyze_metadata

  Trigger metadata analysis to discover keys and generate filter examples. This operation samples documents from the knowledge base, analyzes their metadata fields, updates the key library statistics, and generates new filter examples for few-shot learning. IMPORTANT: This is a potentially long-running operation (1-2 minutes). It samples up to 1000 vectors from the knowledge base and uses LLM analysis to generate filter examples. What happens during analysis: 1. Samples vectors from the Bedrock Knowledge Base 2. Extracts and counts metadata field occurrences 3. Updates the key library with cu…

• chat_with_knowledge_base

  Ask a question and get an AI-generated answer with source citations. Args: query: Your question in natural language (e.g., "What are the API rate limits?") conversation_id: Optional ID to maintain conversation context across multiple queries. Pass the conversation_id from a previous response to continue the conversation. Returns: Multiline string with: - AI-generated answer text - "Sources:" section listing cited documents with titles and URLs (for video/audio sources, includes timestamps like "1:30-2:00") - "[Conversation ID: xxx]" footer for continuing…

• check_key_similarity

  Check if a proposed metadata key is similar to existing keys in the library. Use this before adding documents with new metadata keys to avoid creating duplicate or inconsistent keys. For example, "author" vs "Author" vs "document_author" might all represent the same concept. The similarity check uses fuzzy string matching to find keys that might be duplicates or variations of the proposed key. Args: key_name: The proposed new key name to check (e.g., "author", "doc_type") threshold: Similarity threshold from 0.0 to 1.0 (default: 0.8) - 1.0 = exact match only - 0.8 = …

• generate_image_caption

  Generate an AI caption for an uploaded image using a vision model. This is step 3 (optional) of the image upload workflow. Call this after uploading the image file to S3 but before calling submit_image(). The vision model analyzes the image and generates a descriptive caption that will be used for semantic search in the knowledge base. Args: s3_uri: The S3 URI of the uploaded image (returned by upload_image_url). Format: "s3://bucket-name/path/to/image.jpg" Returns: Multiline string with: - "AI Caption generated!" confirmation - "Caption: <generated_caption>" - The …

• get_configuration

  Get the current RAGStack configuration settings (read-only). Returns all configuration values including chat settings, metadata extraction options, query-time filtering, public access controls, and more. This is read-only - configuration changes require admin access via the dashboard. Configuration Categories: **Chat Settings:** - chat_system_prompt: System prompt controlling AI behavior - chat_primary_model: Primary model for chat (e.g., claude-haiku-4-5) - chat_fallback_model: Fallback when quota exceeded (e.g., nova-micro) - chat_global_quota_daily: Total queries/day for all users - chat…

• get_filter_examples

  Get AI-generated filter examples for metadata-based search queries. Filter examples show how to construct metadata filters for common search patterns. These examples are used as few-shot learning prompts when the system generates filters from natural language queries. Each example includes: - Name: Short identifier for the filter pattern - Description: What the filter does - Use Case: When to apply this filter - Filter JSON: The actual filter syntax (S3 Vectors compatible) Filter Syntax Reference: Basic operators: $eq, $ne, $gt, $gte, $lt, $lte, $in, $nin, $exists Logical operators:…

• get_key_library

  Get the complete metadata key library with all discovered keys. The key library contains all metadata keys that have been extracted from documents across the knowledge base. This is the authoritative list of keys available for filtering and analysis. Unlike get_metadata_stats(), this returns the raw key library data optimized for programmatic use and key suggestions. Returns: Complete list of metadata keys with: - Key name (the field name used in filters) - Data type (string, number, boolean, list) - Occurrence count (document frequency) - Sample values (example values f…

• get_metadata_stats

  Get statistics about metadata keys extracted from documents in the knowledge base. This tool shows what metadata fields have been extracted from your documents, how often each key appears, sample values, and data types. Use this to understand what filters are available for searching. Metadata is extracted during document ingestion using LLM analysis. Common keys include: topic, document_type, date_range, location, author, organization. Returns: Formatted statistics for each metadata key: - Key name and data type (string, number, boolean, list) - Occurrence count (how many docume…

• get_scrape_job_status

  Check the status of a scrape job. Args: job_id: The scrape job ID returned from start_scrape_job (UUID format) Returns: Multiline string with: - "Job: <job_id>" - "URL: <base_url>" - The starting URL - "Title: <page_title>" - Title of the starting page (or "N/A") - "Status: <status>" - One of: PENDING, DISCOVERING, PROCESSING, COMPLETED, FAILED, CANCELLED - "Progress: X/Y pages" - Processed count / total discovered - "Failed: N" - Number of failed pages - Returns "Job <id> not found." if job doesn't exist Errors: - "Error: RAGSTACK_GRAPHQL_ENDPOINT no…

• list_scrape_jobs

  List recent scrape jobs. Args: limit: Maximum number of jobs to return (1-100, default: 10) Returns: Multiline string with: - "Recent scrape jobs:" header - For each job: "[STATUS] title (X/Y pages) - job_id" - Status is one of: PENDING, DISCOVERING, PROCESSING, COMPLETED, FAILED, CANCELLED - Returns "No scrape jobs found." if no jobs exist Errors: - "Error: RAGSTACK_GRAPHQL_ENDPOINT not configured" - Missing endpoint env var - "Error: RAGSTACK_API_KEY not configured" - Missing API key env var - "Error: HTTP error: <details>" - Network failure Example: …

• search_knowledge_base

  Search the RAGStack knowledge base for relevant documents and media. Searches across all indexed content including documents, images, and video/audio transcripts. For media files, results include timestamp information for the matching segment. Args: query: The search query (e.g., "authentication best practices", "what was discussed in the meeting") max_results: Maximum number of results to return (1-100, default: 5) Returns: Multiline string with search results: - "Found N results:" header - For each result: "[index] (score: X.XX) source_path" followed by content snippet…

• start_scrape_job

  Start a web scraping job to add website content to the knowledge base. Args: url: The starting URL to scrape (e.g., "https://docs.example.com/guide") max_pages: Maximum pages to scrape (1-1000, default: 50) max_depth: Maximum link depth to follow from start URL (0-10, default: 3). 0 = only the starting page, 1 = start page + direct links, etc. scope: How far to crawl from the starting URL: - "SUBPAGES" - Only URLs under the starting path (e.g., /docs/*) - "HOSTNAME" - All pages on the same subdomain (default) - "DOMAIN" - All subdomains of the d…

• submit_image

  Finalize an image upload and trigger indexing to the knowledge base. This is the final step of the image upload workflow. Call this after: 1. Getting presigned URL with upload_image_url() 2. Uploading the file to S3 3. Optionally generating AI caption with generate_image_caption() The image will be indexed into the knowledge base using the provided captions for semantic search. At least one caption (caption, user_caption, or ai_caption) should be provided for meaningful search results. Args: image_id: The image ID returned by upload_image_url() (UUID format). caption: Primary captio…

• upload_document_url

  Get a presigned URL to upload a document or media file to the knowledge base. Supported file types: - Documents: PDF, TXT, MD, HTML, DOC, DOCX, CSV, JSON, XML, EML, EPUB, XLSX - Images: JPG, PNG, GIF, WebP, AVIF, BMP, TIFF - Video: MP4, WebM (transcribed via AWS Transcribe) - Audio: MP3, WAV, M4A, OGG, FLAC (transcribed via AWS Transcribe) Video/audio files are automatically transcribed and segmented into 30-second chunks for semantic search. Sources include timestamps for playback. Args: filename: Name of the file to upload with extension (e.g., "report.pdf", "meeting.mp4"). Th…

• upload_image_url

  Get a presigned URL to upload an image to the knowledge base. This is step 1 of the image upload workflow: 1. Call upload_image_url() to get presigned URL and image ID 2. Upload the file to S3 using the returned URL and fields 3. Optionally call generate_image_caption() to get an AI-generated caption 4. Call submit_image() to finalize the upload with captions Supported image types: JPEG, PNG, GIF, WebP, BMP, TIFF Args: filename: Name of the image file with extension (e.g., "photo.jpg", "diagram.png"). The filename determines content type and is displayed in the knowledge base. …

Compare and alternatives

Featured in

• Best MCP servers for knowledge bases and wikis

Signals

Stars

25

Forks

10

Open issues

0

Contributors

5

Weekly downloads

46

Last commit

Oct 21, 2026 (7 days ago)

Last release

Apr 27, 2026 (6 months ago)

Last package publish

Apr 27, 2026 (6 months ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name

io.github.HatmanStack/ragstack

Version

0.1.4

Auth

No auth

Packages

pypi ragstack-mcp

License

Apache-2.0

Repository

github.com/HatmanStack/RAGStack-Lambda

First seen

Oct 24, 2026

Last checked

Oct 24, 2026

Source

Official MCP Registrygithub-graphql