Calyx is a Swift-based browser automation MCP server with 25 well-named tools covering tab management, navigation, DOM interaction, and content extraction. All tools have clear verb_noun naming (browser_open, browser_click, etc.) and descriptions (avg ~80 chars). However, critical gaps undermine production readiness: (1) Input schemas are present but lack explicit type declarations and constraints, parameters are documented in descriptions but not formally constrained (no enums for direction/key values, no min/max for numeric params). (2) Output schemas are completely undocumented, no tool describes what it returns, forcing LLMs to guess. (3) Error handling is absent, no guidance on retryability, user-fixable vs fatal errors, or recovery paths. (4) No security annotations despite write-heavy operations (browser_eval executes arbitrary JS). (5) Transport is unknown (likely STDIO based on Swift/local architecture), capping protocol readiness. The tool set is coherent and well-composed (each does one thing, tab_id chains properly), but lacks the rigor expected of production agent tools.

• Document output schemas for all tools. For browser_list, specify: returns array of {tab_id: string, url: string, title: string}. For browser_get_html, specify: returns {html: string, selector_count: integer}. This enables LLM planning.
• Add enum constraints to free-form parameters: browser_scroll.direction = [up|down|left|right], browser_press.key = [Enter|Tab|Escape|...], browser_wait.visible = boolean. Replace description examples with formal constraints.
• Add error handling guidance to each tool description. E.g., browser_click: 'Returns error if selector matches no elements (try browser_wait first) or element is not visible (try browser_scroll). Retryable: yes.'
• Mark destructive/high-risk tools with security annotations: browser_eval should include a warning that it executes arbitrary code and requires explicit confirmation. Consider adding a dry_run parameter.
• Migrate from STDIO to HTTP+SSE or Streamable HTTP transport to enable remote accessibility and integration with hosted MCP clients. This is required for production agent deployment.