Loading ToolBench

Reading the current corpus…

MakerChecker MCP server · F · ToolBench

MakerChecker

AI agent governance platform with deny-by-default authorization, separation of duties, and hash-chained audit logging for tool call governance

2026-07-28+STREAMABLE_HTTP7 tools

42/100

Definition Quality44

Protocol Readiness15

Supportability57

Spec posture

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

Static source inference · medium confidence · evidence: Streamable HTTP

Current-spec patterns detected

Summary

MakerChecker is a governance/audit framework, not an MCP server. The 7 tools are test fixtures and examples scattered across demo files, not registered in a cohesive MCP server implementation. Tool definitions are inferred from test code rather than explicitly registered. Descriptions are present but minimal (10-50 chars). Input schemas exist but are sparse, most parameters lack type constraints or validation guidance. No output schemas documented. No error handling patterns visible. The codebase focuses on governance wrapping (deny-by-default, audit logging) rather than tool quality. This is a governance SDK, not a production MCP tool server.

Tools (7)

approve-recon@1writeauthsource verified32/100

Approve a reconciliation

csv-ingest@1read onlyauthsource verified47/100

Ingest CSV data from a source file

getQuoteread onlyauth50/100

Read the current market quote for a symbol.

read_fileread onlysource verified52/100

Read a file from disk and return its contents.

report-gen@1writeauthsource verified30/100

Generate a reconciliation report

tidy_workspacedestructivesource verified42/100

Tidy up the workspace.

tradingPushirreversibleauth50/100

Place a live market order with the broker for immediate execution.

Issues (5)

CRITICALtool ↗

No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields.

csv-ingest@1report-gen@1approve-recon@1tradingPushgetQuoteread_filetidy_workspace

HIGHtool-description ↗

Minimal parameter descriptions. 'source' in csv-ingest lacks format/constraint guidance. 'target' in tidy_workspace is vague, does it delete recursively? What if target doesn't exist?

csv-ingest@1tidy_workspace

HIGHconfirmation-request ↗

Irreversible operations (tradingPush, tidy_workspace) lack confirmation or dry-run patterns. No guidance on recovery if agent makes a mistake.

tradingPushtidy_workspace

CRITICALtool ↗

Tool definitions are inferred from test fixtures, not explicitly registered in a cohesive MCP server. No evidence of proper tool registration with capabilities negotiation.

Recommendations

Document output schemas for all tools. E.g., csv-ingest should return {rows: number, columns: string[], preview: object[], errors?: string[]}.
Expand parameter descriptions with constraints. E.g., 'source: Path to CSV file (must exist, max 100MB, UTF-8 encoded)' instead of just 'Path or identifier of the CSV source file'.
Add confirmation/dry-run support to irreversible tools. tradingPush should accept a 'dry_run' boolean; tidy_workspace should return a preview of files to be deleted before executing.
Implement explicit tool registration in a main MCP server entry point, not scattered test fixtures. Use a single tools list with full schemas, descriptions, and error handling.
Add error handling patterns: categorize errors as retryable/user-fixable/fatal, include the invalid value and constraint violated, and suggest recovery steps.

Score history

Overall score trend

First recorded score · v2 rubric

42/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	42	2026-07-28+	v2