Model Context Protocol API Server that creates tools from Bruno API requests
This server has 3 tools with significant quality gaps. The 'echo' and 'list_environments' tools have minimal descriptions (under 20 chars), and the dynamic_bruno_tools tool has a generic description that doesn't explain its purpose clearly. While schemas are present, they lack proper type constraints and are overly generic. The server exposes tool names dynamically from Bruno collections without consistent naming conventions (verb-noun pattern not guaranteed). Error handling is minimal, handlers return basic success/failure objects without actionable recovery guidance. No tool annotations (readOnlyHint, destructiveHint) despite clear WRITE risk on dynamic_bruno_tools. The dynamic tool generation pattern is undocumented and makes parameter validation difficult.
Dynamically created tools from Bruno API requests. Tool names are generated from request names and execute HTTP requests configured in Bruno collections
Echo back a message
List all available environments in the Bruno API collection
Tool descriptions critically short: 'echo' is 'Echo back a message' (17 chars), 'list_environments' is 'List all available environments in the Bruno API collection' (58 chars but lacks context on WHEN to use and WHAT to expect). Both fall below the 50-100 char production baseline.
No tool annotations despite clear risk profiles: 'echo' is READ_ONLY but has no readOnlyHint; 'dynamic_bruno_tools' is WRITE but has no destructiveHint or idempotentHint. This violates the current MCP spec (2026-07-28) which rewards tool annotations. Agents cannot infer whether a tool is safe to call speculatively or retryable.
'dynamic_bruno_tools' is a catch-all that executes arbitrary HTTP requests from Bruno collections. Its name does not follow verb_noun convention (should be something like 'execute_bruno_request'). The description 'Dynamically created tools from Bruno API requests...' is vague and does not explain what the tool does (executes HTTP requests), WHEN to use it (when you have a Bruno request name), or what it returns. Per pattern:tool-description, descriptions must answer these three questions.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 37 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 37 | - | v1 |
Parameters lack actionable descriptions: 'body' in dynamic_bruno_tools has description 'Request body parameters' but does NOT specify format (JSON object? form data?), required fields, or expected structure. 'environment' and 'query' are generic. Current descriptions do not meet this bar.
No output schema documented for any tool. 'echo' presumably returns the input message, but this is not stated. 'list_environments' returns what? A list of environment names? Objects with metadata? 'dynamic_bruno_tools' returns {success, message, status, headers, data} but this structure is not formally documented. Per pattern:tool and mxe:strip-api-responses, tools MUST document output schema so agents know what fields to expect and what data to chain to downstream calls.
'echo' tool violates single-responsibility principle (pattern:tool). Its purpose is unclear, is it a debug/test tool? A utility for transforming text? A placeholder? No clear use case is documented. It should either be removed or renamed and described with a clear purpose (e.g., 'test_mcp_connectivity' with 'Verify MCP server is responding to requests').
Error handling is minimal and not actionable. Handlers return {success: false, message: 'Error: <error message>'}. Per pattern:recovery-guide, errors MUST tell the agent what to do next. Current responses do not categorize errors as retryable, user-fixable, or fatal. An agent gets 'Error: Network timeout' with no guidance on whether to retry or alert the user.
Dynamic tool naming not verb-based. The code uses createToolName(parsedRequest.name) which likely converts 'Create User' → 'create_user' or similar, but the actual naming convention is NOT enforced or documented. If a Bruno collection has a request named 'Do Something', the resulting tool name will be 'do_something', which is a generic verb. Vague verbs like 'do' and 'handle' are explicitly discouraged.
'random_string' parameter in list_environments is a code smell. Its description says 'Dummy parameter for no-parameter tools', this indicates the tool was designed to work around a schema validation issue rather than properly modeling the interface. If list_environments takes no meaningful input, it should have an empty input schema. Dummy parameters confuse agents and waste tokens.