FlareLog catches unhandled errors, silent crashes, and cost spikes in your Cloudflare and AI-built apps — automatically, with one npm install. Then it lets Lovable, Cursor, and Claude read your live logs so your AI can fix them without leaving the editor.
// Free up to 10,000 logs/mo · no card needed · install in under 2 min
Cloudflare's native observability got better in 2025. It still can't tell you when your Worker died silently, when a KV loop is burning money, or why a 500 happened with zero logs. FlareLog can.
When a Worker times out, exceeds its CPU limit, or crashes before your code can log anything, no SDK can report it. The Tail Worker — a small companion Worker Cloudflare runs after each request — still sees the failure.
No boilerplate. The FlareLog SDK picks up console.error and unhandled exceptions on its own, with the context you need to reproduce them.
Cloudflare keeps Workers logs for 7 days at most, and longer retention is still on their roadmap, not shipped. If a customer reports an incident from two weeks ago, the evidence is gone. FlareLog Pro keeps 90 days, searchable.
Click any entry to see the message, source, and trace ID behind it. Example data shown below.
Database connection delay: 450ms
db.d1
tr_218cd98b8df116aa
Connect FlareLog to Claude, Cursor, or Lovable over MCP. Ask "why is checkout failing?" and your assistant reads the real production errors instead of guessing from the code.
Why is checkout returning 500s?
3 errors in the last hour, all TypeError: Cannot read properties of undefined (reading 'plan') from worker.billing at src/billing.ts:88. getUser() returns null when the session has expired, and the handler doesn't check for it. Here's a fix that returns 401 instead of crashing…
{
"mcpServers": {
"flarelog": {
"url": "https://mcp.flarelog.dev",
"headers": {
"Authorization": "Bearer fl_your_api_key_here"
}
}
}
}Your code reports what it knows. A Tail Worker — a companion Worker Cloudflare runs after every request — reports what your code can't: timeouts, CPU limits, and cost.
Embedded inside your standard worker script. Dispatches captured exception payloads and developer info log commands.
console.error / logger.info()Listens out-of-band at the edge interface. Measures the final compute cost, wall CPU, memory limits, and handles silent timeouts.
CPU + class_a_ops + class_b_opsCloudflare's cost metrics lag 15–60 minutes. A misconfigured loop can rack up KV writes, R2 Class A operations, or D1 row reads long before the bill appears. FlareLog estimates the cost of every request in real time from its CPU time, subrequests, and duration.
Immediately calculates costs from incoming Tail Worker CPU & subrequest metrics.
Emails you, and posts to Slack or Discord, when a request's cost crosses your threshold or spikes above your recent average.
Every alert names the Worker responsible, so you know which deploy to roll back.
Install the lightweight SDK package or map our tail listener worker script globally.
Add the SDK in minutes. Get alerted when something breaks or a bill starts to spike — and let your AI fix it without leaving the editor.