Loading ToolBench

Reading the current corpus…

mcp-bridge MCP server · F · ToolBench

← Browse

mcp-bridge

github.com/scmypapa/mcp-bridge

Browser extension and native messaging host for connecting AI web chats to local MCP tools

2026-07-28+STDIO14 tools

39/100

Definition Quality43

Protocol Readiness28

SupportabilityNot measured-

Spec posture

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

Static source inference · medium confidence

Current-spec patterns detected

Summary

MCP Bridge exhibits significant definition quality gaps. Tool naming is inconsistent (mixing verb_noun with colon-separated patterns like 'mcp:get-status'). Most tools lack parameter descriptions entirely, tools/call, resources/read, prompts/get have input schemas but no per-parameter descriptions. Descriptions are present but generic (e.g., 'List all available MCP tools' is 34 chars, below the 50-char baseline for clarity). No output schemas are documented. Error handling is absent, no recovery guidance, no categorization of retryable vs fatal errors. Security concerns: mcp:update-settings accepts an untyped object with no validation or description. The tool registry shows tools are tracked but definitions appear inferred from runtime data rather than explicitly registered with full schemas.

Tools (14)

mcp:get-settingsread onlysource verified35/100

Get current extension settings from local storage

mcp:get-statusread only32/100

Get the connection status of the native host and all configured servers

mcp:get-tab-idread onlysource verified32/100

Get the current tab ID

mcp:get-toolsread onlysource verified43/100

Get all available MCP tools from the native host, with optional force refresh

mcp:open-agent-tabwritesource verified50/100

Open a new tab for agent interaction

mcp:reconnectwrite35/100

Reconnect to the native host and refresh all server connections

mcp:toggle-toolwritesource verified53/100

Enable or disable a specific MCP tool

mcp:update-settingswritesource verified

Issues (5)

CRITICALtool ↗

No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.

tools/listresources/listprompts/listmcp:get-statusmcp:get-tools

CRITICALtool-description ↗

Parameter descriptions missing or incomplete. tools/call, resources/read, prompts/get have input schemas but zero per-parameter documentation. LLMs cannot infer what 'arguments' object should contain.

tools/callresources/readprompts/get

HIGHtool ↗

Inconsistent tool naming convention. Mix of verb_noun (tools/list, resources/read) and colon-separated (mcp:get-status, mcp:toggle-tool) patterns confuses LLM tool selection. Should standardize to one convention.

tools/listmcp:get-statusmcp:get-toolsmcp:toggle-tool

HIGHconstrained-input ↗

Recommendations

Document output schemas for all tools. For tools/list, specify: {type: 'object', properties: {tools: {type: 'array', items: {type: 'object', properties: {name, description, inputSchema, serverName}}}}}.
Add per-parameter descriptions to tools/call, resources/read, prompts/get. E.g., 'arguments: {type: object, description: "Tool-specific parameters. Consult the tool's inputSchema for valid keys and types."}'.
Standardize naming to verb_noun (get_tools, list_resources, toggle_tool) or keep colon-separated but apply consistently across all 14 tools.
Add error handling to tool responses: categorize as retryable (network timeout, rate limit) vs user-fixable (invalid server name, missing parameter) vs fatal. Return actionable messages like 'Server "unknown-server" not found. Available: [list]'.
Constrain mcp:update-settings input. Define allowed keys, types, and ranges. E.g., {type: 'object', properties: {autoRefresh: {type: 'boolean'}, refreshInterval: {type: 'integer', minimum: 1000, maximum: 60000}}, additionalProperties: false}.

Score history

Overall score trend

First recorded score · v2 rubric

39/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	39	2026-07-28+	v2