What problem it solves

Coding agents often need to read many files into their context to locate relevant code and documentation. Projects also need a shared way to search code, structured data, logs, and documents, retain agent memory, and expose those capabilities through familiar APIs and MCP clients.

Key features

• Automatic folder indexing — Point the CLI at a folder and XERJ detects file types, creates typed indices, and extracts code symbols and line numbers through tree-sitter.

  Source: README, engine

• Full-text, vector, and hybrid search — Search indexed content with BM25, vector similarity, or hybrid retrieval through the Elasticsearch-compatible API. Existing Elasticsearch clients, dashboards, and tooling can work against the local server.

  Source: README, docs

• Code-aware retrieval — Retrieve ranked passages with file and line references, and use definitions to jump directly to a symbol and its signature. This supports codebase questions, reference coding, and security-oriented searches.

  Source: README, recipes

• Agent memory and verified facts — Store and recall agent knowledge through namespaced memory APIs. An optional facts model can answer questions from stored facts with cited evidence and the program used to derive the answer.

  Source: README, docs

• MCP and Elasticsearch integration — Expose search through MCP tools for agent clients while retaining an Elasticsearch-compatible HTTP API for standard clients and dashboards.

  Source: README, integrations

Quick start

Copied verbatim from the project's README.

curl -fsSL https://xerj.org/get | sh

xerj --insecure --data-dir ~/xerj-data &      # start it
xerj autoindex ~/my-project              # point it at anything

Who it may suit

Inference: This may suit developers building coding agents, teams that need local code and documentation search, and operators looking for an Elasticsearch-compatible search service that agents can access through MCP.

Topics to discuss

• Which repository and document formats should receive deeper indexing support next?

• How should local search, agent memory, and optional fact-based answering be combined in agent workflows?

• What security and deployment patterns should the community prioritize for shared servers?

Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/xerj-org/xerj#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.