Visual Workflow Automation

Drag-and-drop AI workflow builder. Connect triggers to agents, data sources, and outputs — no orchestration code required.

6 node types

Each node in the flow handles one responsibility. Chain them to build any automation.

Trigger

Starts the flow. Supported: incoming messages (Slack, Discord), cron schedules (hourly, daily, weekly, custom), webhooks from external systems, manual execution, and API calls.

Agent

Sends a prompt to a connected AI agent and waits for the response. The agent has access to its configured connectors and memory. Routes via the nearest edge location.

Connector

Queries any connected data source directly — database query, API call, Slack message fetch, GitHub PR list. Results passed to the next node.

Transform

Processes data between nodes. Format output as JSON or text, filter results by condition, or parse structured data from agent responses.

Escalation

The silent monitoring node. Runs an agent without surfacing output to users. Only escalates — posts to a channel, sends an email, fires a webhook — when a configurable condition is met.

Output

Delivers results. Post to any messaging platform, send an email, return an API response, or fire a webhook to any external system.

The Escalation Node

Build monitoring workflows that stay quiet. The escalation node runs an agent silently on every trigger — incoming message, scheduled check, webhook event — and only surfaces output to humans when a configurable condition is met.

Use expr-eval conditions: severity > 2, count > 100, status != "ok"

Example flows

Intelligent incident triage

Webhook → Parse → Agent (triage) → Escalation (severity > 2) → Slack alert + PagerDuty

Incoming alerts are parsed, triaged by an AI agent, and only escalated to humans when severity warrants it.

Scheduled data briefing

Schedule (9am daily) → DB query → Agent (summarize) → Slack message

Every morning, query your database, have an AI summarize the results, and post to a Slack channel.

Multi-agent research pipeline

Webhook → Agent A (gather) → Transform → Agent B (analyze) → Output (API)

Chain multiple agents together — one gathers data, another analyzes it, results returned via API.

Full execution history

Every flow execution is recorded with per-node timing, retry history, and complete output at each step.

120s

Flow timeout

90s

Agent node timeout

Configurable retries