Herd: FLOW · AI-native workflow & agentic orchestration

Every app’s tools.
One intelligent flow.

Herd: FLOW is a drop-in AI workflow component that discovers API and MCP tooling across your applications, pairs them with persistent company context, and orchestrates multi-step work with human guardrails.

MCP discovery, multi-app tooling, persistent data catalog, AI chat & voice agent, human-in-the-loop approval forms, intelligent step retries, full logging, node-level cost accounting, and one-click reusable workflows.

Watch the demo Scroll the steps

Herd: FLOW component showing Hub and Spoke multi-agent DAG, Architecture Review cost breakdown, and Workflow Overview

The demo · 3:38 · with voiceover

From API & MCP discovery to a reusable team workflow.

Walk through the embeddable Herd: FLOW component: discovering 65 MCP and API capabilities, loading persistent datasets for people, brand and products, building an SLT meeting workflow via AI Chat and the built-in Voice Conversation Agent, enforcing human approval guardrails, adapting on retry, tracking node and workflow costs, and turning a proven run into a generic template for the whole team.

Captured in the Herd: FLOW workflow engine component. Captions available. The soundtrack in the header pauses while the video plays.

Step by step · 12 capabilities

Scroll the flow engine.

Every capability of Herd: FLOW, one screen at a time. Each image comes directly from the component.

01Component & Graph

One drop-in component. Any application.

The entire Herd: FLOW engine ships as a single embeddable component with an imperative control API, 55+ typed events, and interactive DAG visualization across Force, Tree, Circular, Grid, Radial, and Custom layouts.
02MCP & API Discovery

Discover and authenticate with AI guidance

Point Herd: FLOW at any MCP server or REST API. It performs the JSON-RPC 2.0 handshake, negotiates authentication headers, encrypts credentials with AES-256-GCM, and uses AI to categorize and test all 65 capabilities.
03Multi-App Tooling

Tools from every app, working together

Instead of isolated point integrations, the engine understands tool schemas across email, SMS, voicemail, calendar, CRM, and company transcripts simultaneously — chaining 50+ tools to get real business work done.
04Dataset Catalog

Persistent context for People, Brand and Products

Live tooling is preferred when data changes by the minute. For steady context — your employee directory (HERD Employees), Herd AI Brand Guidelines, and product catalog (HERD Catalog) — the persistent Data Library gives every workflow shared context and saves API cost.
05AI Chat & Reasoning

Create and run new workflows via AI Chat

State your goal in plain language: set up an SLT meeting for today at 3:00 PM EST. The Reasoning Agent decomposes the goal into steps, wires up context datasets like HERD Employees, and executes the pipeline.
06Voice & Approval Forms

Speech-to-Text, Text-to-Speech, and human approval

Click the Herd elephant in the bottom-right corner to talk to the built-in Conversation Agent with Speech-to-Text and Text-to-Speech — no typing required. When a step needs confirmation, Herd: FLOW generates an interactive approval form on the fly.
07Human Guardrails

Nothing consequential happens without sign-off

Every AI-generated form renders in live Preview Mode with field validation, version history, and privacy controls — preventing the engine from sending emails, creating records, or triggering charges without human approval.
08Triggers & Constraints

Schedules, events, and six built-in safety rules

Launch workflows on a cron schedule, a database event, or an API call. Six built-in constraints and custom Approval, Warn, and Stop gates enforce governance alongside a real-time LLM token trace.
09Intelligent Retry

Intelligent AI retry when steps hit a snag

With configurable step retries and AI failure diagnosis, the engine adapts queries rather than failing. Here, the verification step bypassed an unparseable transcript field, matched attendee emails via HERD Employees, and confirmed all 8 calendar records.
10Cost Accounting

Full cost accounting at the node and workflow level

Before execution, Architecture Review estimates LLM calls and dollar cost per node — 16 LLM calls across 6 nodes totaling $0.0140 — and enforces a hard workflow cost cap ($5.00) during execution.
11Full Logging

Every run, every cost, and every guardrail receipt

The Runs dashboard logs duration, step counts, exact workflow spend ($0.2792, $0.1166), and plain-English guardrail audit trails — including why the booking tool refused to double-book Eric Marquez during a 3:00 PM conflict.
12Reusable Workflows

Proven workflows become generic for every user

Once a workflow succeeds, click Make reusable. Herd: FLOW extracts personal details like Matt Francis and specific attendee names into pre-run input questions, turning a one-off run into a shared Schedule a Team Meeting template.