Sift Releases is an MCP server tool by Deths74r, built with Python 3.11+, designed to transform website content into a verifiable corpus for AI agents. Its sift-cli component offers prebuilt binaries for various operating systems and a Python package. This system allows AI agents to consume current, self-hosted information, ensuring provable information retrieval where answers trace back to exact sources. Sift Releases helps AI agents access and process web data efficiently, supporting data management by focusing on changed content.

What does Sift Releases connect to?

The Sift Releases MCP server connects to any reachable website on the internet. It indexes the content from these sites, converting the raw data into a structured, verifiable corpus. This corpus is specifically designed for consumption by AI agents. The server operates within the Model Context Protocol (MCP) framework, acting as a central point for AI systems to gather and process external web-based information.

Core features and tools

Sift Releases exposes several callable tools for managing its operations. These include list_users, list_report_templates, and check_for_updates. Access tiers define further capabilities: the 'create' tier allows actions like create_*, upload_dataset, and append_test_measurements. The 'destructive' tier grants access to tools such as update_*, archive_*, and unarchive_*, enabling modifications and data lifecycle management within the verifiable corpus.

Who is it for?

Sift Releases targets AI agents and their developers or users who require current, verifiable, and self-hosted information sourced directly from websites. It is particularly valuable for "grep-first AI agents" that need to trace information back to its original source, content hash, and snapshot. The system supports a need for provable information retrieval. This is not the right fit if your AI agents do not require verifiable data from external websites.

How to get started

Access is granted through a standard OAuth authorization flow. For remote access, generating TLS certificates and a bearer token is required, passed via an "Authorization: Bearer <token>" header over streamable HTTP transport.

Use cases

• Indexing any website into a verifiable corpus for AI agents to consume via MCP.
• Enabling provable information retrieval, where AI agent answers can be traced back to the exact source, content hash, and snapshot.
• Supporting incremental and low-operations data management by re-extracting only changed content and re-deriving data from cached raw sources.

Sift Releases provides a specialized MCP server for AI agents to consume verifiable web content. Its tiered access controls and support for streamable HTTP make it suitable for secure, traceable information gathering. The system is built to support AI agents needing to ground their responses in provable, self-hosted data from websites. Try indexing a website to create a verifiable corpus for an AI agent.