Loading ToolBench
A local CLI-based AI developer assistant that uses Ollama and DeepSeek Coder to explain code, search for patterns, debug files, suggest improvements, and answer coding questions without requiring internet connectivity.
This server has 12 tools with basic schemas and descriptions, but significant quality gaps prevent production use. All tools have descriptions (10-100 chars, mostly adequate), and input schemas are present with type definitions. However, parameter descriptions are sparse or missing entirely (e.g., 'recursive' in list_files has no description), output schemas are completely undocumented, and error handling provides no recovery guidance. No tool annotations (readOnlyHint, destructiveHint) despite all being read-only. The server is a CLI wrapper around Ollama with no MCP protocol implementation visible, it appears to be a standalone Python application, not an MCP server.
Answer general coding questions with optional code context.
Help debug a piece of code by analyzing it and an error message.
Explain what a piece of code does with improved prompting.
Find class or type definitions by name.
Find function definitions by name in various languages (Python, JavaScript, Java, C#, C++, Go, Rust).
Explain how a function is used in given code.
Get metadata about a file.
No output schemas documented. Tools return unstructured responses (e.g., 'explanation' string from explain_code). LLMs cannot plan downstream calls or extract structured data without knowing response fields.
Parameter descriptions incomplete or missing. 'recursive' in list_files has no description; 'case_sensitive' in search_in_file lacks guidance on default behavior. LLMs cannot infer parameter semantics from names alone.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). All 12 tools are read-only but lack explicit annotation. Agents cannot determine safety/retry semantics without annotations.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 39 | 2026-07-28+ | v2 |
List all files in a directory with Windows-friendly path handling.
Read file content with Windows-safe path handling.
Search for a query across all valid files in the project.
Search for a query in a single file.
Suggest improvements for a piece of code focusing on quality, readability, performance, best practices, and security.
Error handling provides no recovery guidance. No evidence of try-catch blocks, validation, or actionable error messages. If a file is not found or regex fails, agents receive no guidance on next steps.
Not an MCP server. Source code is a standalone CLI application (cli.py, ollama_client.py) with no MCP protocol implementation, transport layer, or tool registration mechanism visible. Cannot be used as an MCP server without complete rewrite.