Loading ToolBench

Reading the current corpus…

toolbx MCP server · F · ToolBench

toolbx

Local MCP gateway — configure your MCP servers once and connect every MCP client (Claude, Codex, OpenCode) to one place.

<=2025-11-25STDIO11 tools

45/100

Definition Quality48

Protocol Readiness47

Supportability38

Spec posture

Inferred effective spec <=2025-11-25·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging

Deprecated protocol patterns detected

Summary

Toolbx exposes 11 tools with mixed quality. Most tools have descriptions (9/11), but parameter documentation is sparse and inconsistent. Tool names follow verb_noun convention (tool import, tool list, tools search), which is good. However, schemas are incomplete: many parameters lack type information or descriptions. The 'greet' tool is a test fixture and should not be exposed. Error handling guidance is absent. Output schemas are not documented. No tool annotations (readOnlyHint, destructiveHint) despite clear risk levels (DESTRUCTIVE for tool remove, WRITE for tool enable/disable). Parameter descriptions are often generic ('override the resolved config path') without explaining constraints or expected formats.

Tools (11)

greetread only50/100

Greets the named person.

tool disablewrite50/100

Disable an imported custom tool.

tool enablewrite50/100

Enable an imported custom tool.

tool importwrite50/100

Import a local .ts or .js custom tool file into Toolbx.

tool inspectread only50/100

Show a custom tool manifest plus a head of its source file.

tool listread only50/100

List imported custom tools.

tool removedestructive50/100

Remove an imported custom tool (deletes its source file and manifest entry).

tools disablewrite50/100

Disable an upstream tool from being exposed through Toolbx.

tools enablewrite50/100

Enable an upstream tool exposed through Toolbx.

Issues (5)

CRITICALconfirmation-request ↗

Destructive tool (tool remove) lacks confirmation pattern and error recovery guidance. No dry-run or confirmation step before deleting source files and manifest entries.

tool remove

HIGHtool-annotations ↗

No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels declared in metadata. LLMs cannot infer which tools are safe to retry or have side effects.

tool importtool removetool enabletool disabletools enabletools disable

HIGHconstrained-input ↗

Parameter descriptions are generic and lack constraints. 'config' parameter appears in 10 tools with identical description ('override the resolved config path') but no guidance on format, precedence, or validation rules.

tool importtool inspecttool listtool removetool enabletool disabletools listtools searchtools enable

Recommendations

Add tool annotations to all tools: mark 'tool remove' as destructiveHint=true, 'tool import' and 'tool enable/disable' as idempotentHint=true, and read-only tools with readOnlyHint=true. This enables LLMs to reason about retry safety and side effects.
Document output schemas for all tools. For 'tools list' and 'tools search', specify: returns array of {name, description, server, enabled} objects; for 'tool inspect', returns {manifest, sourcePreview}. Include field types and whether results are paginated.
Enhance parameter descriptions with constraints and examples. Replace 'override the resolved config path' with 'Path to TOML/JSON config file (e.g., ~/.toolbx/config.toml). If omitted, uses default search paths.' Add format/pattern constraints for 'exposedName' (e.g., 'namespace__tool_name format').
Implement confirmation pattern for 'tool remove': add a 'confirm' boolean parameter (default false) or return a confirmation token that must be passed in a second call. Prevent accidental deletion of tool source files.
Remove 'greet' test fixture from production tool registration. Move it to a test-only server or exclude it from the tool list via a filter.

Spec posture evidence

Inferred effective spec: <=2025-11-25.

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

Score history

Overall score trend

First recorded score · v2 rubric

45/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	45	<=2025-11-25	v2