Loading ToolBench

Reading the current corpus…

mcp-proxy MCP server · C · ToolBench

← Browse

mcp-proxy

github.com/sparfenyuk/mcp-proxy

A MCP server which proxies requests to a remote MCP server over streamable HTTP or SSE.

2026-07-28+STREAMABLE_HTTP11 tools

64/100

Definition Quality61

Protocol Readiness72

Supportability62

Spec posture

Inferred effective spec 2026-07-28+·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging, Dynamic Client Registration

Deprecated protocol patterns detected

Summary

The server provides 11 well-named tools with clear action verbs and consistent naming conventions (list_*, get_*, read_*, call_*, subscribe_*, unsubscribe_*, set_*, complete). All tools have short, actionable descriptions (15-85 chars). However, parameter descriptions are minimal or absent for several tools, and output schemas are not documented in the visible code. The server proxies remote MCP capabilities, which is a valid but limited architectural pattern. Parameter schemas are present for most tools but lack richness (no enums, ranges, or detailed constraints). Error handling appears basic based on the available source.

Tools (11)

call_toolirreversible50/100

Call a tool on the remote MCP server with support for progress forwarding

completeread only50/100

Request completion/autocomplete from the remote MCP server

get_promptread only50/100

Get a specific prompt from the remote MCP server

list_promptsread only50/100

List available prompts from the remote MCP server

list_resource_templatesread only50/100

List resource templates from the remote MCP server

list_resourcesread only50/100

List available resources from the remote MCP server

list_toolsread only50/100

List available tools from the remote MCP server

read_resourceread only50/100

Read a resource from the remote MCP server

set_logging_levelwrite

Issues (6)

HIGHtool ↗

Output schemas not documented. No visible response type specifications for any tool. Users cannot determine what fields to expect from list_prompts, read_resource, call_tool, etc.

list_promptsget_promptlist_resourceslist_resource_templatesread_resourceset_logging_levelsubscribe_resourceunsubscribe_resourcelist_toolscall_toolcomplete

HIGHtool-description ↗

Parameter descriptions are minimal or absent. 'set_logging_level' takes a 'level' param with no constraint documentation (valid values unclear). 'complete' has two params ('ref', 'argument') with generic descriptions that do not explain format or expected structure.

set_logging_levelcomplete

MEDIUMconstrained-input ↗

No enum constraints on single-value parameters. 'set_logging_level' accepts a string 'level' but does not declare valid values (DEBUG, INFO, WARNING, ERROR, CRITICAL) as an enum. LLM will guess or hallucinate invalid levels.

Recommendations

Document output schemas for all 11 tools. For list_* tools, specify the structure of returned items, include a 'total' or 'has_more' field, and document pagination if supported.
Add enum constraints to 'set_logging_level.level' parameter: enum: [DEBUG, INFO, WARNING, ERROR, CRITICAL].
Expand parameter descriptions: 'set_logging_level.level' should state 'The logging level to set on the remote MCP server. Must be one of: DEBUG, INFO, WARNING, ERROR, CRITICAL.'
Clarify 'complete' parameters: rename 'ref' to 'completion_ref' and 'argument' to 'argument_to_complete'. Add descriptions explaining the expected format (e.g., 'The argument object being completed, with structure depending on the tool/prompt being completed').
Add pagination support: for list_prompts, list_resources, list_tools, add optional 'limit' (default 20, max 100) and 'offset' parameters. Return a 'total_count' field in responses.
Implement actionable error responses. When a remote server is unreachable, return: 'Remote MCP server not responding. Check server status and network connectivity.' When a tool does not exist on the remote server, return: 'Tool "<name>" not found on remote server. Call list_tools to see available tools.'
Add tool annotations (readOnlyHint, destructiveHint) to indicate side effects. Mark subscribe_resource, unsubscribe_resource, set_logging_level with destructiveHint or writeHint as appropriate.
Document the meta.progressToken parameter in call_tool: 'Optional progress token for forwarding progress updates. If provided, progress messages from the remote tool will be relayed to the client.'

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Logging (deprecated) - log to stderr or use OpenTelemetry

Score history

Overall score trend

↑ 22 points across a rubric change (v1 → v2)

64/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	C	64	2026-07-28+	v2
2026-03-09	F	42	-	v1