Skip to main content

Documentation Index

Fetch the complete documentation index at: https://raindrop.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

Raindrop Triage is an investigation agent for your AI app. Ask it anything you’d ask a teammate who lives in your data: root causes, user impact, rollout status. It pulls the relevant events, finds the pattern, and replies with the conversations that prove it. It runs in three places:
  • Slack: @Raindrop from any channel and continue in-thread.
  • Web: chat with Triage from inside the Raindrop app.
  • MCP: your coding agent (Claude Code, Cursor, Codex, etc.) can call it too. See the MCP overview.

​
Ask Triage

@mention Raindrop with a question. Triage runs the investigation, replies in thread with examples and a hypothesis, and keeps context across follow-ups (“is this still happening?”, “didn’t we fix it last week?”) so you can keep poking until you understand the issue.
@mention Raindrop in Slack
A few things to try:
  • @raindrop how's the rollout on flag checkout-v2?
  • @raindrop is this still happening? (with a screenshot of a customer complaint)
  • @raindrop what are the top issues affecting enterprise customers this week?
Triage drives the same tools you’d reach for in the app (semantic and regex search over events, signal counts, trace inspection, charts, experiments) and grounds every answer in real conversations you can click into.

​
Agent Briefs (scheduled investigations)

Agent Briefs are scheduled investigations Triage runs on a cadence and posts back to Slack. Describe what you want investigated and when, and Triage saves the prompt, the schedule, and the destination, then runs the investigation against fresh data on every tick.
Agent brief delivered to Slack
Two ways to create one:
@mention Raindrop with a scheduling request. Triage will confirm and save it.
@raindrop every Monday at 9am, summarize the biggest issues enterprise customers had last week
Prefixing with Remind me: also works:
@raindrop Remind me: every weekday at 8am, surface tool error spikes from the past 24h
What’s configurable:
  • Prompt: the investigation Triage runs each tick (e.g. “summarize tool error rates each morning”)
  • Schedule: any cron expression. Common cadences like daily, weekly, or weekday-only are inferred from natural language
  • Timezone: defaults to your org’s default timezone
  • Channel: Slack channel to post to (or your DM)
  • Status: active or paused, plus a Run now action
Common briefs teams set up:
  • Summarize tool error rates each morning
  • Weekly, show me which users churned last week and what issues they hit
  • Every Monday, highlight any new use cases from last week
  • Every day until Friday, report on the checkout-v2 rollout

​
Daily Digest

There’s also a built-in Daily Digest that ships out of the box: a morning summary of patterns and changes from the day before. Toggle it on at Settings → Notifications. Compact mode posts one-liners with details in-thread.

​
Custom Monitors

Set thresholds on any signal (error rate, frustration rate, a custom classifier) and Raindrop pings the channel when the threshold is crossed. Configure them at Settings → Notifications under Signal Alerts, or ask Triage in Slack to set one up. See Alerts for the full configuration options.
Custom monitors

​
Create signals from chat

When an investigation surfaces a pattern worth tracking going forward, ask Triage to track it and it’ll draft a signal for you, no leaving the conversation. 
@raindrop track every time the assistant apologizes for not being able to help
Triage scopes the behavior, opens a Deep Search labeling artifact, and seeds it with examples it just found. You label a few matches in the app, and Raindrop turns the labels into a saved signal that runs over every future event (and backfills the prior 3 days). Triage picks the right signal type for you:
  • Regex signals: for literal text patterns RE2 can capture (specific phrases, error strings, URLs, response templates).
  • Classifier signals: for semantic concepts that aren’t lexical (frustration, confusion, refusal style, intent).
You can refine in chat too: “only count negative mentions”, “tighten this to LinkedIn URLs only”, and Triage adjusts the proposal and re-samples. See Signals for how custom signals work end-to-end.

​
In your coding agent

Triage is also available as an MCP server, so Claude Code, Cursor, Codex, and any other MCP-compatible coding agent can call it directly. Your agent pulls a live issue from production, reads the failing examples, edits the relevant file, runs tests, and opens a PR, all without leaving your editor. 
> claude "check raindrop for the top enterprise issue and fix it"

→ Calling raindrop...
→ Found: auth token expiration not handled during SSO refresh
  (43 conversations affected, started April 22)
→ Reading src/auth/tokenRefresh.ts
→ Editing src/auth/tokenRefresh.ts, added retry logic for expired SSO tokens
→ Running tests... ✓ all passing
→ Opening PR #847
See the MCP overview for setup instructions for each editor.

​
Get started

1

Install

Add Raindrop to Slack with one-click OAuth. No infra to run.
2

Pick channels

Choose where Triage posts proactive alerts, monitors, and Agent Briefs at Settings → Notifications.
3

Ask

@mention Raindrop with a question, or schedule an Agent Brief.
For coding-agent setup, see the MCP overview.

​
Privacy

For information on how Raindrop collects, uses, and protects your data, see our Privacy Policy.
Raindrop uses AI to generate issue analyses, digests, and conversational responses. AI-generated content may be inaccurate, incomplete, or misleading. Always verify critical information independently before making decisions based on Raindrop’s outputs. Triage is a tool to assist your team, not a replacement for human judgment.