preflight-mcp
# Preflight MCP Server
TrustPilot for APIs, built for AI agents. Independent reliability ratings for every API and MCP server — powered by synthetic probes and crowdsourced agent telemetry.
## Install
```bash
pip install preflight-mcp
```
## Usage
### As a standalone MCP server
```bash
preflight-mcp
```
### With Claude Desktop
Add to your Claude Desktop MCP config:
```json
{
"mcpServers": {
"preflight": {
"command": "preflight-mcp"
}
}
}
```
### With Claude Code
```bash
/plugin install preflight@claude-plugins-official
```
## Tools
### check_reliability
Look up the independent reliability rating for any service.
```
check_reliability(service="openai-api")
```
### compare_services
Compare ratings side by side to pick the most reliable provider.
```
compare_services(services=["openai-api", "anthropic-api"])
```
### report_outcome
Leave a review — report what happened when you called a service. Improves ratings for everyone.
```
report_outcome(service="openai-api", outcome="success", latency_ms=200)
```
## Configuration
| Variable | Default | Description |
|----------|---------|-------------|
| `PREFLIGHT_API_URL` | `https://preflight.beamt.io` | API base URL |
| `PREFLIGHT_API_KEY` | (none) | API key for authenticated requests |
Get a free API key at [beamt.io/preflight](https://beamt.io/preflight).
## License
MIT
TDQS
Scored across 3 tools
Each tool has a clearly distinct purpose: check_reliability evaluates a single service, compare_services compares multiple services side-by-side, and report_outcome submits user feedback. There is no overlap in functionality—an agent would never confuse which tool to use for a given task.
All tools follow a consistent verb_noun pattern with clear, descriptive names: check_reliability, compare_services, and report_outcome. The naming is uniform and predictable, making it easy for agents to understand each tool's function at a glance.
With only 3 tools, the server feels slightly thin for its reliability-checking domain. While the tools cover core functions, additional utilities like listing top-rated services or fetching historical trends might be expected. However, the count is not severely inadequate.
The toolset covers the essential reliability workflow: checking individual services, comparing them, and contributing data. A minor gap is the lack of a tool to fetch aggregated reliability data (e.g., a leaderboard or trending services), but agents can work around this by using compare_services with a predefined list.