# Definable Workflow — Multi-agent AI workflows that build, run, and self-heal

> Describe the work in plain English. Three orchestrating agents — Builder, Executor, Verifier — coordinate a team of specialists (Researcher, Writer, Designer, Engineer, Distributor) across 50+ models and 1,000+ apps. Self-healing by design. No flowcharts. No glue code.

## What Workflow is

Workflow is the multi-agent automation system at the centre of Definable. The user types an outcome in plain English. The Builder agent reads the prompt, plans the steps, and picks the right specialists and models. The Executor agent runs every step. The Verifier agent checks every output and self-heals failed steps automatically.

The user describes; the agents build, run, and verify. The result is a shipped outcome — not a draft.

## How a Workflow runs — the three-agent loop

Every Definable workflow is orchestrated by three agents:

1. **Builder agent** — reads the plain-English prompt, plans the workflow steps, identifies which specialist agents are needed, and assigns the best model per step. Typically GPT-5.5.
2. **Executor agent** — runs each step in sequence or in parallel where possible. Calls APIs, invokes specialists, executes tool calls across the user's 1,000+ connected apps.
3. **Verifier agent** — checks every output. When a step fails (rate limit, schema drift, off-brand response, malformed data), the Verifier rejects it and re-routes through the Executor with corrective context. The retry runs automatically.

## The specialist agents

The Builder assembles a team from these specialists for each workflow:

- **Researcher (GPT-5.5)** — scans sources, pulls competitor context, synthesises findings, briefs the team
- **Writer (Claude Opus 4.7)** — brand-voice drafts, long-form posts, email sequences, copy
- **Designer (Studio)** — cinematic photos, character-consistent storyboards, product shots, social-ready video
- **Engineer (Def Code CLI)** — plans, builds, tests, debugs against the user's repo conventions
- **Distributor (1,000+ apps)** — schedules, posts, syncs, sends across every tool in the user's stack

## How it differs from Zapier, Make, n8n, Lindy, Gumloop

Other automation tools require the user to build workflows themselves — node by node, drag-and-drop, glue code between nodes. They execute every step with brittle, pre-written code; when an API shifts shape, the flow breaks and someone has to fix it.

Workflow is different. The Builder agent builds the workflow from a plain-English prompt. The Verifier agent monitors every step at runtime, detects failures, and retries until the work completes. Resilient. Self-correcting. Zero maintenance.

## How it differs from ChatGPT, Claude, Perplexity

Chatbots give the user answers — drafts, text, summaries. The user still has to take the actions.

Definable agents take the actions across 1,000+ apps. The Verifier loop ensures the output is correct before it reaches the user. Workflow ships the work, not a draft.

## Capabilities

- **Plain-English workflow creation** — describe outcomes, not steps
- **Multi-agent specialist team** — Researcher, Writer, Designer, Engineer, Distributor
- **Multi-model routing per step** — best model assigned to each agent role automatically
- **Self-healing Verifier loop** — rate limits, schema drift, off-brand responses auto-retried
- **Multi-step orchestration** — chain unlimited steps with no per-task pricing
- **Branching and conditions** — if/then/else with AI reasoning between steps
- **Parallel fan-outs** — run independent steps simultaneously
- **Event triggers** — workflows fire on webhooks, schedules, app events
- **Cross-tool reasoning** — agents interpret each step's output and decide what to do next
- **Audit-ready run logs** — every step replayable, every change tracked
- **1,000+ app integrations** — CRM, marketing, communication, e-commerce, analytics, dev tools
- **No per-task pricing, no per-app limits** — included in every plan

## Worked example: campaign launch

Prompt: "Launch a product campaign on LinkedIn, X, and Instagram. Use our brand voice."

- **Step 1 — Researcher** (GPT-5.5) scans 14 competitor posts. Identifies the angle. Done. Verifier passes.
- **Step 2 — Writer** (Claude Opus 4.7) drafts the campaign copy. Verifier rejects: tone drifted off-brand. Executor re-runs with corrective context. Verifier passes the self-healed draft.
- **Step 3 — Designer** (Studio) generates hero image plus 3 social assets. Verifier passes.
- **Step 4 — Distributor** schedules across LinkedIn, X, Buffer. Verifier passes.
- **Result**: Campaign live in 6m 38s. 1 self-heal. 0 manual fixes.

## Use cases

- **Marketing** — campaign launches, content series, social distribution
- **Sales** — lead enrichment, scoring, CRM hygiene, follow-up sequences
- **Customer success** — onboarding flows, churn prevention, NPS follow-ups
- **Operations** — invoice processing, expense management, vendor workflows
- **Engineering** — incident response, deploy notifications, monitoring automation
- **E-commerce** — order fulfillment, customer service routing, review collection
- **Research** — competitor analysis, market scans, briefing reports

## Pricing

Workflow is included in every Definable plan — no per-task pricing, no per-app limits.

- Starter (free): 5,000 credits/month, all integrations enabled
- Creator ($19/month): 75,000 credits/month, pro models, unlimited workflows
- Pro ($49/seat/month): 250,000 credits/seat, team workspaces, advanced Verifier
- Enterprise (custom): SSO, SCIM, audit logs, dedicated infrastructure, signed BAA

India pricing from ₹399/month.

## Related

- HTML page: https://definable.ai/workflow/
- Assistant: https://definable.ai/assistant/
- Knowledge Bases: https://definable.ai/knowledge/
- Pricing: https://definable.ai/pricing/
- Compare to Zapier: https://definable.ai/compare/definable-vs-zapier/
- App catalog: https://definable.ai/apps/
- Glossary — What is workflow automation?: https://definable.ai/glossary/workflow-automation/
- Use case — Replace Zapier with AI: https://definable.ai/use-cases/replace-zapier-with-ai/