Loading ToolBench

Reading the current corpus…

platypus MCP server · C · ToolBench

← Browse

platypus

github.com/willdady/platypus

REST backend for Platypus - a multi-agent chat platform with plugin support, sandbox execution, and MCP integration

2026-07-28+STREAMABLE_HTTP5 tools

67/100

Definition Quality71

Protocol Readiness56

Supportability67

Spec posture

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

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

Deprecated protocol patterns detected

Summary

Platypus Backend provides 5 well-named sandbox tools with clear, comprehensive descriptions and complete input schemas. All tools follow verb_noun naming convention (shellExec, fsRead, fsWrite, fsEdit, fsList). Descriptions are concise (50-150 chars), explain WHEN to use each tool, and document limitations (output capping, truncation). All parameters have type definitions and descriptions. Risk annotations (IRREVERSIBLE, READ_ONLY, WRITE) are present and actionable. Output schemas are not explicitly documented in the source, which is a notable gap for production tooling. Parameters use appropriate constraints (enums for mode, lineRange with format hints). Error handling is mentioned in descriptions (truncated flag, not-found cases) but recovery guidance in error responses is not visible in the provided code.

Tools (5)

fsEditwrite50/100

Edit a file by replacing exactly one occurrence of `oldString` with `newString`. Fails if `oldString` is absent or appears more than once. Prefer this over rewriting whole files.

fsListread only50/100

List files and directories in the sandbox. Pass `recursive: true` to descend, and `glob` (e.g. "**/*.ts") to filter. Output is capped; check the `truncated` flag.

fsReadread only50/100

Read a UTF-8 text file from the sandbox. Paths are relative to the workspace root. Pass `lineRange: [start, end]` (1-indexed, inclusive) to read a slice. Large files are truncated.

fsWritewrite50/100

Write a UTF-8 text file in the sandbox. `mode: "create"` fails if the file already exists; `mode: "overwrite"` replaces it. Parent directories are created automatically.

shellExecirreversible50/100

Run a shell command in the sandbox. Each call starts a fresh shell — there is no persistent shell state between calls. Pass `cwd` (relative to the workspace root) to choose the working directory. Output is capped; check the `truncated` flag.

Issues (4)

HIGHtool ↗

Output schemas not explicitly documented. The rubric requires documentation of return types so LLMs know what fields to expect for chaining. While risk classifications (IRREVERSIBLE, READ_ONLY, WRITE) are present and helpful, formal schemas for tool outputs (e.g., shellExec returns {stdout, stderr, exitCode, truncated}) are not visible in the source.

shellExecfsReadfsWritefsEditfsList

MEDIUMrecovery-guide ↗

Error recovery guidance not visible. Descriptions mention output capping and truncation flags, but the source does not show how tools guide LLMs on error cases (e.g., 'If output is truncated, call with a more specific command'). Error responses should include actionable next steps.

shellExecfsList

MEDIUMconstrained-input ↗

Missing parameter constraint documentation for some fields. The 'env' parameter in shellExec is documented as 'Environment variables to pass' but lacks format guidance (JSON object structure, allowed keys, value types). Similarly, 'glob' in fsList should document glob syntax conventions.

Recommendations

Add explicit output schemas for all 5 tools. Example: 'shellExec returns {stdout: string, stderr: string, exitCode: number, truncated: boolean, truncatedReason?: string}'.
Document recovery paths in error scenarios. E.g., 'If output is truncated (truncated: true), try: 1) request only specific fields, 2) add a limit flag, or 3) split the command into smaller steps.'
Clarify 'env' parameter format: 'Environment variables as a JSON object, e.g., {"PATH": "/usr/bin", "DEBUG": "1"}. Keys are case-sensitive; values must be strings.'
Document glob pattern syntax for fsList: 'Glob patterns follow standard Unix conventions (*, ?, [abc], **). Examples: **/*.ts (all TypeScript files), src/**/*.test.ts (test files under src).'
For shellExec, add optional 'dryRun' parameter: 'If true, print the command without executing it. Useful for validating commands before destructive operations.'
Add per-parameter type refinements in schema. For example, 'lineRange' should specify 'array of exactly 2 positive integers: [startLine, endLine], 1-indexed and inclusive.' Currently generic array type.

Spec posture evidence

Inferred effective spec: 2026-07-28+.

Relies on Sampling (deprecated) - integrate directly with the LLM provider API

Score history

Overall score trend

↑ 44 points across a rubric change (v1 → v2)

67/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-22	C	67	2026-07-28+	v2
2026-03-09	F	23	-	v1