MCP Server Checker
Validate MCP servers for protocol compliance, code quality, best practices, and LLM ergonomics.
Features
- Protocol Compliance: Validates JSON-RPC 2.0 format, handshake sequence, capability negotiation
- Functional Testing: Tool discovery, tool execution, error handling
- Maturity Scoring: Two-score system (Compliance + Maturity) with badge levels
- Multiple Profiles: Lenient, Standard, Strict, Paranoid validation levels
Installation```bash
cd /path/to/mcp-server-checker
python -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
pip install -e ".[dev]"
## Quick Start
### As MCP Server
Add to your `.mcp.json`:```json
{
"mcp-server-checker": {
"command": "bash",
"args": [
"-c",
"cd /path/to/mcp-server-checker && .venv/bin/python -m mcp_server_checker.server"
]
}
}
Using the Tools```python
Check a server
result = await check_server(
server_path="python -m my_mcp.server",
profile="standard"
)
List available checks
checks = list_checks()
Get validation profiles
profiles = get_profiles()
## Validation Profiles
| Profile | Maturity Minimum | Use Case |
|---------|------------------|----------|
| Lenient | 50% | Early development |
| Standard | 70% | Production servers |
| Strict | 85% | High-quality servers |
| Paranoid | 95% | Maximum compliance |
## Badge Levels
| Badge | Requirements |
|-------|--------------|
| CERTIFIED | Compliance PASS, Maturity 95%+ |
| PRODUCTION_READY | Compliance PASS, Maturity 85%+ |
| FUNCTIONAL | Compliance PASS, Maturity 70%+ |
| DEVELOPMENT | Compliance PASS, Maturity 50%+ |
| NON_COMPLIANT | Compliance FAIL |
## MVP Phases - ALL COMPLETE
All 9 MVPs are implemented:
- **MVP-1**: Protocol + functional (handshake, tool calls, basic errors)
- **MVP-2**: Transport + lifecycle (stdout corruption, shutdown, cancel, ping)
- **MVP-3**: Utilities (logging, progress, pagination)
- **MVP-4**: Content + contract (schema-runtime tests)
- **MVP-5**: Code analysis (structure, docs, patterns)
- **MVP-6**: Security (secrets, injection, path traversal)
- **MVP-7**: Host compatibility (Claude Desktop, env startup)
- **MVP-8**: LLM ergonomics (naming, descriptions, schema)
- **MVP-9**: Scoring + fix packs (weighted scoring, remediation)
## Fix Packs
Remediation suggestions grouped by effort:
| Pack | Effort | Focus |
|------|--------|-------|
| docs_baseline | ~30 min | Documentation |
| schema_hardening | ~1 hr | Tool schemas |
| error_hygiene | ~2 hr | Error handling |
| security_baseline | ~2 hr | Security |
| lifecycle_compliance | ~3 hr | MCP lifecycle |
| llm_polish | ~1 hr | LLM ergonomics |
| transport_hardening | ~1 hr | Transport layer |
## Running Tests```bash
pytest tests/ -v # 173 tests
Specification
Based on MCP Protocol Specification 2025-11-25.
See KB/project-background-specificationV2.md for the full specification document.
Showcase
See the arbiter showcase site for interactive demos and documentation.
License
MIT
