Provides nine hash-verified file operations that prove the AI agent is working against current content, not stale context. Each line carries a content hash that must match before edits apply, preventing corruption from mid-conversation file changes. Exposes read_file_with_hashes, apply_diff, insert_lines, delete_lines, and validation hooks that let agents confirm their view is fresh. Sits inside a broader skill marketplace that treats it as optional fallback: skills work without MCP but use these tools when available to save tokens and catch drift. Designed for long coding sessions where the file might change between tool calls, whether from external edits or multi-agent workflows.
claude mcp add --transport stdio io.github.levnikolaevich-hex-line-mcp uvx hex-line-mcp