Framework for building Model Context Protocol (MCP) servers in TypeScript
mcp-framework is a framework for building MCP servers, not itself a complete server with production-grade tool definitions. The 5 example tools show inconsistent quality: secure_data and echo are minimal proofs-of-concept; docs tools (get_page, list_sections, search_docs) are well-structured but lack output schema documentation. Naming follows verb_noun convention. Descriptions are present but vary in completeness (10-170 chars). Schemas are visible in source but parameter descriptions and output schemas are not consistently documented. Error handling is basic and lacks recovery guidance. Security considerations are partially addressed (OAuth example) but not enforced across all tools.
Echoes the input message back
Retrieve the full content of a documentation page by its slug or URL path. Returns the page as markdown.
List the documentation structure showing all sections and their page counts. Use this to discover what documentation is available before searching.
Search the documentation by keyword or phrase. Returns a ranked list of matching pages with relevant excerpts.
Query secure data (requires OAuth authentication)
Output schemas not documented for any tool. LLMs cannot infer what fields to expect in responses, forcing them to guess at downstream data access patterns.
echo tool is too generic and lacks clear purpose. Description ('Echoes the input message back') is minimal (40 chars, below 50-char baseline). Tool names like 'echo' and 'process' do not convey actionable intent to LLMs.
secure_data tool lacks permission/scope declarations. No documentation of what OAuth scopes are required, which principals can invoke it, or what audit trail is maintained.
Error handling is minimal across all tools. get_page returns a string ('Page not found: ...') but no structured error with recovery guidance. No clear distinction between retryable and user-fixable errors.
Inferred effective spec: 2025-06-18+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 65 | 2025-06-18+ | v2 |
| 2026-03-09 | F | 37 | - | v1 |
search_docs accepts limit (1-25) but lacks documented pagination strategy. No 'next_cursor' or 'total_count' returned in visible output schema, forcing agents to guess how to iterate large result sets.
Parameter descriptions for 'section' (optional filter) lack examples of valid section names. LLMs cannot determine what sections exist without calling list_sections first, but discovery relationship is not documented.
get_page tool implementation is truncated in source (line ends with '.replac'), making full schema inspection impossible. Cannot verify complete parameter or output handling.