agents don't complain.
they silently fail_
a broken tool call shouldn’t end the job. problems brings failures to your team and puts known fixes back in the agent’s hands.
need a hand getting started? .
see what’s happening inside your tools.
All sources All operations Last 15 minutes
Errors
3 groups · 44 errors| Operation / error | Occurrences | Source | Last seen |
|---|---|---|---|
| create_charge | 26 | MCP | |
| send_receipt | 12 | CLI | |
| list_customers | 6 | MCP |
understand usage
see which tools and commands agents call, and how often.
spot repeated failures
similar errors come together, so recurring problems stand out.
inspect the details
open a group to see the error and its recent occurrences.
add to your MCP or CLI in minutes
npm install @problemsdev/sdk
Wrap your server before registering tools. Calls and errors are batched automatically.
import { createProblems } from "@problemsdev/sdk";
const problems = createProblems({ publicKey: "YOUR_PUBLIC_KEY" });
problems.wrapMcpServer(server);
// Register tools as usual, after wrapping the server.
// Successful calls and errors are batched automatically.Record a named command’s outcome, then flush before exiting.
import { createProblems } from "@problemsdev/sdk";
const problems = createProblems({ publicKey: "YOUR_PUBLIC_KEY" });
try {
await runDeploy();
problems.recordCommand("deploy", { exitCode: 0 });
} catch (error) {
problems.recordCommand("deploy", { exitCode: 1, error });
process.exitCode = 1;
} finally {
await problems.flush();
}agents hitting problems?
grab 25 minutes with our founder. tell us where your agents get stuck and see your tool activity live.
