Three live AI workflow modes

Pick a workflow, run it, inspect the operational trace.

This is a working agent console for business research and sales ops. Each workflow has a different job: audit a website, enrich a CRM lead, or draft the next follow-up.

Run one of the 3 workflows

The tabs are the product. Start by choosing the job you want done, then use sample data or your own input.

Prospect research

Audit a company site for weak positioning, missing proof, and useful outreach angles.

No run yetChoose a workflow above, then run it. Clean results appear here, not raw JSON.

Run trace

WaitingNo run yet.

Architecture

The proof is the operating shape: inputs, tools, model calls, fallback paths, traceability, and API contracts.

01
Input contractValidate JSON bodies and reject missing fields early.
02
Tool layerScrape pages, clean HTML, extract links and contact signals.
03
Model adapterWorkers AI wrapper with fallback model chain and JSON extraction.
04
Workflow logicEach agent owns decisions, warnings, and deterministic fallback behaviour.
05
Proof envelopeEvery response returns output, metadata, warnings, and a run trace.

API contracts

The UI uses the same endpoints documented in the README, so the demo and repo stay aligned.

POST /api/prospect-research

{ "url": "https://company.com" }

POST /api/crm-enrichment

{ "company": "Acme", "url": "https://acme.com" }

POST /api/follow-up-draft

{ "thread": "Previous email context..." }