XERJ Brings Local AI Search and MCP Server Capabilities to Coding Agents

Overview

XERJ is a local-first search engine and Model Context Protocol (MCP) server designed specifically for AI agents and developers. Distributed as a single Rust binary, it aims to reduce token consumption by indexing local codebases, documentation, logs, and structured files like CSV, JSON, and PDFs, exposing an Elasticsearch-compatible API.

According to project documentation, the primary use case is reference coding—allowing coding agents to query indexed open-source repositories or internal codebases to retrieve specific implementation patterns rather than reading entire files or relying solely on grep.

Core Features and Capabilities

• Elasticsearch-Compatible API: Integrates with standard client libraries and dashboards without requiring a Java runtime.
• Automated Indexing: Uses tree-sitter for AST-aware code parsing, extracting symbols and line numbers from source files.
• Multiple Search Modalities: Supports BM25 full-text search, vector embeddings, and hybrid search.
• Agent Memory: Implements endpoints such as `/_memory/{namespace}` for persistent agent long-term memory and recall.
• MCP Integration: Acts as an MCP server via `xerj mcp`, exposing tools for search, definition lookup, and repository mapping to supported desktop assistants and CLI tools.

Measured Performance and Claims

Project documentation cites controlled study results showing that reference coding via XERJ reduced agent output tokens significantly compared to grep-based approaches or memory alone on specific benchmark tasks. However, these benefits depend on the dataset; the documentation notes that retrieval provides little advantage for popular public libraries that a model has already memorized.

Security and Configuration

By default, XERJ binds to `127.0.0.1` for local development. The `--insecure` flag disables TLS and API-key authentication, granting all callers superuser privileges over the local instance. Production or shared environments require proper API key configuration (`--api-key` or `XERJ_API_KEY`) to prevent unauthorized access.

How to Verify

Users can inspect the official repository, review the security model documentation, test local installation scripts, and examine client configuration recipes on the project's website.