We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/grafana/mcp-grafana'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
README.md•1.23 KiB
# Tests
This directory contains an e2e test suite for the Grafana MCP server.
The test suite evaluates the LLM's ability to use Grafana MCP tools effectively:
- **Loki tests**: Evaluates how well the LLM can use Grafana tools to:
- Navigate and use available tools
- Make appropriate tool calls
- Process and present the results in a meaningful way
- Evaluating the LLM responses using `langevals` package, using custom LLM-as-a-Judge approach.
The tests are run against two LLM models:
- GPT-4
- Claude 3.5 Sonnet
Tests are using [`uv`] to manage dependencies. Install uv following the instructions for your platform.
## Prerequisites
- Docker installed and running on your system
- Docker containers for the test environment must be started before running tests
## Setup
1. Create a virtual environment and install the dependencies:
```bash
uv sync --all-groups
```
2. Create a `.env` file with your API keys:
```env
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
```
3. Start the required Docker containers
4. Start the MCP server in SSE mode; from the root of the project:
```bash
go run ./cmd/mcp-grafana -t sse
```
5. Run the tests:
```bash
uv run pytest
```
[`uv`]: https://docs.astral.sh/uv/