Anysearch

•

Empowers AI agents with specialized, domain-filtered information search, research, and knowledge retrieval capabilities.

Anysearch is a command-line interface (CLI) and MCP server designed as a vertical-domain information-specialist agent. It integrates advanced search, research, memory, and knowledge functionalities, featuring FTS5 for efficient recall and multi-source Reciprocal Rank Fusion (RRF) for combining results. A core capability is its robust domain allowlisting, which strictly gates returned information both at the provider level and post-retrieval, ensuring an 'abstain-first' approach rather than hallucination when no allowed results are found. It also includes an auto-indexing MCP server that provides tools for various agent hosts.

主要功能

01Abstain-first policy for zero gated results, providing a first-class abstain outcome instead of errors or hallucinations.
02Multi-source search with RRF fusion across providers and FTS5 for memory recall.
03MCP server offering tools like `search_web`, `research_web`, `recall_memory`, `query_knowledge`, and `ans_chat`.
04Verified integration with major agent hosts including CodeBuddy, Claude, Codex, and Antigravity.
05Vertical-domain information specialization with authoritative URL allowlists for strict content gating.
060 GitHub stars

使用案例

01Integrating advanced search and research capabilities into existing agent host platforms like CodeBuddy or Claude.
02Developing AI agents that require precise, domain-specific information retrieval and adherence to content policies.
03Ensuring strict information governance by filtering all search and research results against a predefined domain allowlist.