Five minutes to your first execution.

Create an account at mesedi.vercel.app/signup. A default project is created for you. The Hobby tier is free and gives you enough executions to evaluate Mesedi against a real workload.

Open the dashboard, head to API Keys, and click Create key. The raw key is shown once, copy it to your password manager or your shell environment immediately. The dashboard only ever stores the prefix and a hash; we cannot recover the raw key for you later.

export MESEDI_API_KEY="mesedi_sk_..."

Pick the one that matches your stack.

Python

pip install mesedi

TypeScript / Node

npm install mesedi

A wrapped function becomes an observed execution. Mesedi records start, completion (or crash), wall-clock duration, and emits a stable crash signature so identical exceptions cluster into a single failure group instead of paging you a hundred times.

Python

import os
import mesedi

mesedi.configure(api_key=os.environ["MESEDI_API_KEY"])

@mesedi.wrap
def run_my_agent(query: str) -> str:
    # your agent logic
    return "answer"

run_my_agent("hello")

TypeScript

import { configure, wrap } from "mesedi";

configure({ apiKey: process.env.MESEDI_API_KEY! });

const runMyAgent = wrap(async (query: string) => {
  // your agent logic
  return "answer";
});

await runMyAgent("hello");

Head back to mesedi.vercel.app/app. The execution you just ran appears under Executions within a few seconds. Click in to see the timeline of LLM calls, tool calls, validators, and any events that landed during the run.

If a failure-class detector fires (crashes, time budget, step count, tool failures, validator failures, prompt injection, cost velocity, drift, or a loop), the dashboard groups it under Failure groups with a one-screen playbook explaining what the signal means and what to try first.

Python SDK reference covers @tool, AsyncShipper, Anthropic auto-instrumentation, hard-halt with budgets, and the LangChain / CrewAI adapters.

TypeScript SDK reference covers tool(), AsyncShipper, Anthropic patching, hard-halt + SSE remote channel, and the Vercel AI SDK adapter.

Failure classes and playbooks walks through each of the seven detector signatures and what each Tier 1 Playbook recommends.