The platform

One platform, two ways to deploy it

PhantomOps isn't one product — it's an umbrella. Agent personas are full digital employees that own an inbox and a workflow end to end. Point solutions are focused capabilities you can run standalone or embed inside any agent. Mix and match to what you actually need.

Agent personas

Full digital employees

Each agent runs as its own identity — its own mailbox, its own machine, its own memory — and operates entirely through email, so your team adopts it with zero new UI to learn.

Sally — Sales & Solutioning

Sally is PhantomOps' first agent, live in production today. She reads RFPs and tenders, drafts proposals and Pega-aligned solution documents, builds decks, spreadsheets, letters and reports, schedules meetings, and recommends staffing based on past project history.

  • ✓ Reads & writes PDF, Word, Excel, CSV & PowerPoint — with OCR
  • ✓ Books Outlook meetings and delivers files to network folders
  • ✓ Web research, only when a task actually needs it
  • ✓ Runs across four LLM providers from one agent
Mark — Marketing & Brand

Mark is the creative engine — generating social and campaign content, brand-consistent imagery, and proposal banners. His skills are now delivered as toggleable capabilities inside Sally, so one agent can carry both a sales and a marketing voice.

  • ✓ Social copy & on-brand graphics
  • ✓ Proposal banners & cover imagery
  • ✓ Success-story synthesis from project data
Critic — Quality Gate

Before any draft reaches a manager, the Critic runs a hidden second-pass review — checking for contradictions, missing sections, or unsupported claims — and attaches a confidence score. It's a quality filter, not a substitute for human sign-off.

  • ✓ Contradiction & gap detection
  • ✓ Confidence scoring on every draft
  • ✓ Source attribution enforced on outputs
Noor — Arabic Localization

Noor is the regional differentiator — native Arabic communication and content, purpose-built for GCC customers and their regulators, not a translated afterthought bolted onto a global product. Noor is built on the Arabic NLP Processor below — the same engine, packaged as a full persona.

  • ✓ Native Arabic drafting & customer replies
  • ✓ On-brand, compliant tone
  • ✓ Purpose-built for the GCC market
Point solutions

Focused capabilities, not a full agent commitment

Every point solution below started as internal tooling built to make an agent safer or sharper — now productized so you can license and deploy each one standalone, or embed it inside any PhantomOps agent.

PI Guard

Strips secrets and personal information from a request before it ever reaches an external LLM — the same privacy scrubber that runs in front of every PhantomOps agent call, now available as its own service.

  • ✓ Redacts PII, credentials & forbidden terms pre-call
  • ✓ Drop-in gateway in front of any LLM API
  • ✓ Direct answer to CBUAE data-residency guidance
Magic Reader

An OCR engine tuned for frail, aged and low-legibility documents — faded ink, poor scans, handwritten margins — the kind of paperwork that trips up generic OCR but is common in long-running BFSI archives.

  • ✓ Built for degraded scans, not clean PDFs
  • ✓ Feeds structured text straight into any workflow
  • ✓ Strong fit for KYC & legacy-document digitisation
Arabic NLP Processor

A dedicated Arabic language engine for customer-service teams — understands dialect and formal Arabic, drafts and classifies replies, and powers Noor's persona when you want a full agent instead of a component.

  • ✓ Dialect-aware, not machine-translated
  • ✓ Plugs into an existing service desk or CRM
  • ✓ The engine underneath the Noor persona
How it works

One email in, finished work out

No new UI, no login to learn. Managers interact with agent personas exactly like they would a colleague — by email — with a human-in-the-loop review cycle at every step. Point solutions integrate differently: as an API call in front of your existing LLM traffic (PI Guard), a document-processing endpoint (Magic Reader), or an NLP service wired into your service desk (Arabic NLP Processor).

1
Assign
Email the agent a task — draft this proposal, summarise this document, schedule this meeting.
2
Gather & draft
The agent reads any attachments, searches only if the task needs it, and drafts the output.
3
Quality gate
The Critic reviews the draft for gaps or unsupported claims and attaches a confidence score.
4
Review & refine
You receive the finished file by email. Reply with changes — the agent keeps full thread context.
Aligned, not locked in

Model-agnostic, by design

A single LLM gateway routes each task to the right model: local models for triage and routine drafting, premium reasoning models for high-stakes work. Built on open standards — LangGraph orchestration and the Model Context Protocol — so we ride the frontier instead of betting on one vendor.

  • • Cheap models first, reasoning models when it counts
  • • Swap or add a provider without re-architecting
  • • No customer lock-in to a single hyperscaler
Claude Gemini Copilot ChatGPT Perplexity Llama · local

Provider marks above are stylized identifiers, not the providers' registered logos.

Free tier

Experiment with PhantomOps, hands-on

We publish free PhantomOps versions for developers, enthusiasts and prospective customers to try — a capped, low-cost local mode, self-serve, with community skills. It doubles as a living demo of the platform.

  • • Self-serve setup, no sales call required
  • • Local-model routing keeps it free to run
  • • A real, hands-on look at the platform before you buy

The free tier is in active rollout. Leave your details and we'll notify you the moment self-serve access opens.

Join the free-tier waitlist

Submitting opens a pre-filled email in your own mail client — nothing is sent to a third party without your review.

Need something more than the free tier?

Talk to us about a governed pilot on your own infrastructure.

Request a BFSI pilot