TryCrucible — Prove your AI engineering skills

Scenario

Your MCP server must implement 5 tools: list_directory, read_file, write_file, search_files, and delete_file. It will be tested against 15 tool call sequences, including multi-step operations and edge cases like non-existent paths and empty directories.

Expected output

All 15 test tool call sequences return correct responses including proper error handling

Dataset

📦

Dataset files

Scoring rubric

Clarity5%

MCP manifest is complete, README explains how to connect a client

LLM usage20%

Tool descriptions are precise and unambiguous for the model to use correctly

Robustness10%

Handles invalid paths, permission errors, and concurrent calls gracefully

Correctness25%

All 15 test sequences return expected output, error codes match spec

Architecture20%

Clean MCP protocol implementation, tools are well-typed and documented

Language-free evaluation

Build your solution in any language or framework — Python, TypeScript, Go, Rust, Java, C#, or anything else. The dataset artifacts may be in one language; your implementation does not need to match. TryCrucible evaluates the behaviour of your system, the quality of your AI workflow, your verification strategy, and the reproducibility of your submission — not your language choice.

Submission requirements

A public GitHub repository link
A Dockerfile in the repo root — any language or framework; the evaluator builds and runs your container
Your solution reads test_inputs.json from the working directory and writes results.json — standard I/O contract across all challenges
A decisions.md — 3–5 sentences on the key architectural and AI-workflow choices you made
The system must be fully reproducible — we clone, build, and run it against real test inputs

Evaluation contract

When you submit, the evaluator runs these steps in order:

1Clone your public GitHub repository
2Build your container from the Dockerfile in the repo root
3Mount test_inputs.json into the working directory
4Run your solution in a network-isolated sandbox (5 min limit, 512 MB RAM)
5Read results.json from the working directory
6Score correctness against hidden ground truth, then score architecture, AI workflow, robustness, and clarity

// I/O contract — same schema for every challenge

Input (provided by evaluator)

// test_inputs.json
[
  { "id": "t1", "input": { ... } },
  { "id": "t2", "input": { ... } }
]

Output (written by your solution)

Create a free account to start

Already have one? Sign in

Build an MCP server for filesystem access