Preflight

The TrustPilot for APIs, built for AI agents.

Independent reliability ratings for every API and MCP server — powered by synthetic probes and community reviews from real agent traffic. Know which services to trust before your agent calls them.

Blind calls burn money

Every time your agent calls an unreliable API, you pay for the tokens that built the request, the retry logic, the fallback handling, and the user’s patience. Your agent doesn’t know that API had 12 incidents last quarter. Preflight does.

Status pages lie

“All Systems Operational” while your agent gets 500s. Status pages update 15–45 minutes after an outage starts — if they update at all. Preflight runs its own synthetic probes every 5 minutes. No self-reporting. No lag. Real numbers.

MCP servers are the Wild West

Your agent connects to an MCP server. Is it up? Does it respond in under a second? Nobody tracks this. There’s no Datadog for MCP. Preflight monitors MCP servers with protocol-level handshake probes — not just HTTP pings.

Look up a rating before every API call.

Your agent looks up the rating. If it’s below your threshold, route to a fallback. No wasted tokens. No failed tasks. No angry users.

Request

GET https://preflight.beamt.io/v1/score?service=openai-api

Response

{
  "score":   { "overall": 94 },
  "status":  "operational",
  "trend":   "stable",
  "metrics": {
    "uptime_7d":      99.7,
    "latency_p50_ms": 230,
    "error_rate_7d":  0.003
  }
}

What you get

Compare ratings before you commit

Which LLM provider is most reliable right now? Not last month — right now. Get a side-by-side comparison with a recommendation and the reason behind it.

Community-powered, tamper-proof

Like TrustPilot, agents that opt in leave reviews after every call. More agents reviewing means more accurate ratings. Trust weighting prevents any single source from gaming the numbers.

MCP-native

Preflight itself is an MCP server. Your agent can check reliability, compare services, and leave reviews without writing HTTP code.

MCP Tools

check_reliability("openai-api")
compare_services(["anthropic-api", "openai-api"])
report_outcome("stripe-api", "success", latency_ms=140)

Set up in under a minute

Pick the method that fits your stack. All roads lead to the same reliability ratings.

Claude Code

Plugin

/plugin marketplace add https://gitlab.com/beamt/preflight-mcp.git
/plugin install preflight@beamt-preflight-mcp

Preflight tools are available to Claude automatically. No configuration needed.

Claude Desktop

MCP Config

Add to claude_desktop_config.json: 

{
  "mcpServers": {
    "preflight": {
      "command": "preflight-mcp"
    }
  }
}

Requires preflight-mcp to be installed first (see pip below).

pip

Any MCP Client

pip install preflight-mcp

Run standalone:

preflight-mcp

Works with any MCP-compatible client — Cursor, Windsurf, custom agents.

Direct API

No MCP needed

curl https://preflight.beamt.io/v1/score?service=openai-api \
  -H "Authorization: Bearer YOUR_API_KEY"

No install needed. Get a free API key from the signup form above.

Optional: Set an API key

Without a key, Preflight tools work but return limited data (free tier, overall rating only). To unlock detailed ratings and higher limits, set your API key: 

export PREFLIGHT_API_KEY=pf_live_your_key_here

Or in your MCP config:

{
  "mcpServers": {
    "preflight": {
      "command": "preflight-mcp",
      "env": {
        "PREFLIGHT_API_KEY": "pf_live_your_key_here"
      }
    }
  }
}

Who this is for

AI agent developers building agents that call external APIs and MCP servers in production. If your agent makes decisions based on third-party service availability, you need Preflight.

Agent frameworks

LangChain, CrewAI, AutoGen, custom agents

MCP client developers

Claude Desktop, Cursor, Windsurf, custom MCP clients

Platform teams

Running fleets of agents that depend on external services

Pricing

Free

$0

/month

  • 5,000 queries/month
  • Overall rating
Popular

Builder

$9

/month

  • 100,000 queries/month
  • Detailed ratings (uptime, latency, reliability)
  • Community reviews

Scale

$29

/month

  • 1,000,000 queries/month
  • Detailed ratings (uptime, latency, reliability)
  • Community reviews

Fleet

$79

/month

  • 10,000,000 queries/month
  • Detailed ratings (uptime, latency, reliability)
  • Community reviews

Get started in 10 seconds

Get an API key instantly. No credit card. No sales call. 

curl -X POST https://preflight.beamt.io/v1/keys \
  -H "Content-Type: application/json" \
  -d '{"email": "you@company.com"}'

You get back a key. Start looking up ratings immediately.

You wouldn’t book a hotel without checking reviews. Why would your AI agent call an API without checking its rating? Preflight is TrustPilot for APIs — independent reliability ratings so your agent knows which services to trust. The alternative is finding out the hard way.