Skip to main content
Raindrop home dashboard
AI agents fail differently than traditional software. A 500 error is obvious. An agent that confidently gives wrong information, forgets what a user said 3 messages ago, or takes a bad path fails silently, invisible to traditional monitoring. Raindrop helps you:
  • Visualize agent trajectories: see every tool call, error, and recovery in seconds
  • Detect issues automatically: choose from presets that catch forgetting, user frustration, and task failures
  • Search across millions of interactions: find issues in production data with natural language
  • Track custom signals: define any signal you care about (syntax errors, aesthetic complaints, an agent stuck in a loop) and track it at scale
  • A/B test your agents: run experiments to validate that fixes actually worked

​
Quickstart

Run this in your project directory:
It installs the raindrop CLI and runs raindrop cloud setup, which:
  1. Signs you in to Raindrop in the browser.
  2. Writes your org write key to ./.env as RAINDROP_WRITE_KEY.
  3. Installs the raindrop-setup and raindrop-investigate skills and the hosted MCP server into your coding agents (Claude Code, Cursor, Codex, and others).
Then run /raindrop-setup in your coding agent. It finds your AI call sites, instruments them, and checks that events reach Raindrop.
To send events to a specific project, append --project <slug>. Setup then prints a prompt to paste into your coding agent in place of /raindrop-setup, and the skill writes the slug into your SDK call. The Get started page in the dashboard gives you this command with the project already filled in.
Prefer to wire it up yourself? Pick an SDK or integration below.

​
Integrate Raindrop

TypeScript

Full-featured SDK with tracing support for Node.js and edge runtimes.

Python

Native Python SDK for FastAPI, Django, and other Python frameworks.

HTTP API

RESTful API for any language or platform.

Integrations

Connect Raindrop with AI frameworks, cloud providers, and developer tools.

​
Core Features

Signals

Ground truth indicators for agent performance. Track wins, failures, and custom behaviors across all your interactions.

Triage Agent

An agent that investigates your AI agents in Slack, on the web, and as an MCP.

Issues

Automatic detection of broader distribution changes worth investigating.

Experiments

A/B test your agents to validate changes. Compare models, prompts, and configurations.