First waves shipping now · Full vision Q3 2026 · Waitlist open

One command. Every job done.

TRD Pilot turns 641 specialist agents into one autonomous worker for your business. Tell it what you want — in plain language, on the channel you already use. It figures out the steps, does the work across your site, your messaging, and the platforms you use every day, and brings it back to you done.

You
Launch tomorrow's promotion across the site, WhatsApp regulars, and Instagram.
Pilot coordinating · 4 agents engaged · awaiting your approval before publish
Built on: TRD Platform · Compute · Storage · OS Global from launch: US · Europe · MENA · Asia · ANZ Languages: 10 today, 18+ Q4 2026 Industries: 87 live, 107 target Build time: ~35s prompt → site
TRD PILOT · AI BRAIN · LIVE

One command in. Every agent out.

A plain-language command lands on the Pilot Brain. The orchestrator parses intent, dispatches the right specialists from the 641-agent fleet, runs the work across your site / messaging / booking / delivery / social, and surfaces one signed diff for one-tap approval.

ORCHESTRATOR · ROUTING
87 INDUSTRIES · 641 AGENTS
PILOT BRAIN
641
agents · live
💬
WhatsApp
🗨️
Telegram
📲
SMS · Voice
🍽️
Toast · Square
📅
Booking
📸
Instagram · IG
🛒
💳
📧
🗓️
641
Industry agents
87
Industries live
5
Global regions
Signed receipts
Why TRD Pilot

Copilot helps. TRD Pilot does.

Every autonomous-agent product shipping today was built for either developers or large enterprises with implementation teams. None of them is built for the small business running your favourite restaurant, salon, retail store, fitness studio, clinic, or shop. That's the gap TRD Pilot is built to fill.

The current options

Horizontal. Enterprise-priced. US-first.

Generic web-browsing agents from $200/mo and up. Aimed at power users and developers. Don't know the difference between updating a restaurant menu and a salon booking page. Don't speak the channels your customers actually use. Don't run end to end — they assist, you finish.

TRD Pilot

Vertical. SMB-priced. Global from day one.

641 industry-specialist agents that already know how restaurants, salons, retail, professional services, fitness, real estate, hospitality, education, photography, construction, and 80 other industries actually work. One command. Multi-channel. End-to-end execution with a full audit trail. Priced where small-business software has always lived — not where enterprise IT lives.

The current options
TRD Pilot
Who it's built for
Developers, power users, enterprise IT teams
Small-business owners across 87 live industries (on the way to 107) — no IT department needed
Pricing
$200–$249/month per user
Sits in the same tier as TRD Platform — small-business prices
Availability
US-only or US-first; some still invite-only
Five regions in parallel — US, Europe, MENA, Asia, ANZ
Industry knowledge
Generic — one agent that tries to do everything
641 specialist agents across 87 live industries; each one knows its domain
Interface
A browser tab; sometimes a desktop app
Messaging-first — WhatsApp, Telegram, IG, SMS, email, voice, web
What it operates
A browser, generically — figures out each site from scratch
Your site + messaging + booking + delivery + social — in lockstep
Audit trail
A history log; some screenshots; not portable
Signed receipts pinned to TRD Storage — tamper-evident, permanent, yours
Failure handling
Often fails silently; you check later
Never silent — pauses, screenshots, asks you on your channel, then continues
Owns the underlying stack
No — sits on top of someone else's models and infra
Yes — TRD owns the Platform, Compute, Storage, and OS layers
How it works

From one sentence to done.

Three steps. You don't see the work — you see the outcome, with a clear receipt of everything the agents did so you can audit, undo, or repeat any of it in one click.

01

Tell TRD Pilot what you want

Plain language. No menus to learn, no workflow builder to figure out. Send it from wherever you already are — the dashboard, WhatsApp, Telegram, SMS, even by replying to an email.

WebWhatsAppTelegramSMSEmailVoice
02

The right agents collaborate

The orchestrator picks specialists from 641 agents trained across 87 live industries — the ones that actually know your business. They coordinate, do the work, ask one clarifying question if they need it, and keep going.

87 industries641 agentsMulti-channel
03

You get the outcome, with proof

Pilot returns a clean summary: what changed, where, when, and how much it cost. Every action emits a signed receipt pinned to TRD Storage. Undo any step. Repeat the whole sequence with one click.

Signed receiptsOne-click undoReusable
Use cases by industry

Six businesses, six commands.

These are not hypotheticals. Each one is built from the actual tools that industry uses, the actual channels customers come through, and the actual agents in TRD's fleet. Owner sends one message; Pilot does the rest.

Restaurant

Launch tomorrow's specials everywhere at once

Pasta night tomorrow — three new specials, 20% off for regulars before 7pm, push it everywhere we sell.

The restaurant agent drafts the special-menu page; the photo agent generates plate visuals matching brand style; the menu-sync agent updates Toast, Square, the website, and the delivery-app listing; the WhatsApp agent broadcasts the regulars list with the early-bird code; the Instagram agent posts a story and a feed image with the offer; the booking agent opens an extra 7pm reservation slot.

Owner gets a clean diff: 11 changes across 6 surfaces, 4 minutes, awaiting approval. One tap publishes everything.

4 mins prep6 surfaces0 copy-pastesigned receipt
Salon & beauty

Recover a quiet Wednesday before it ruins the week

Wednesday's bookings are way down again — figure out why and fix it.

The salon analytics agent notes Wednesday bookings dropped 28% vs the four-week average; the local research agent flags that a nearby competitor opened a Wednesday discount; the campaign agent drafts a midweek-treat promotion targeted at clients who haven't visited in 6+ weeks; the SMS agent prepares the broadcast; the calendar agent reshapes Wednesday's slots toward the most popular services.

Owner sees the diagnosis, the proposed plan, and the projected lift — +22% expected based on past similar promotions — and approves with one tap.

auto-diagnosedpast data usedprojected liftone approval
Retail & e-commerce

Run a flash sale across every channel without missing one

Flash sale at 6pm — 30% off the autumn collection, two hours only. Coordinate everything.

The retail merchandising agent selects the autumn-collection SKUs and stages the price changes; the storefront agent prepares a sale banner and updates the homepage hero; the Shopify agent stages the discount codes and inventory holds; the marketplace agent prepares the same pricing on connected channels; the social agent drafts six platform-native posts; the email agent prepares the broadcast.

Everything is scheduled to go live at 6pm and revert at 8pm exactly. Owner sees the full plan and gets a 2-hour countdown timer with one rollback button if anything goes wrong.

scheduled dropauto-revertone rollbackcross-channel
Fitness & studios

Fill an empty class slot 90 minutes before it starts

7am pilates is half-empty tomorrow — fill it.

The fitness CRM agent pulls members who've taken 7am pilates before, filters for those who haven't booked this week; the messaging agent drafts a personalized invite with a one-tap booking link; the WhatsApp + SMS agents send to opt-in channels respecting each member's preference; the waitlist agent surfaces three drop-ins from the past week who didn't get a spot last time.

Owner sees: 14 invites sent, 6 spots remaining, projected fill rate 80%. The class books up by 6:45am. Pilot returns a summary: who came, who didn't, who hadn't been seen in 3+ weeks for a retention follow-up.

targeted inviteschannel preferencefollow-up loop
Clinic & healthcare

Confirm tomorrow's bookings and rescue the no-shows

Confirm tomorrow's appointments and offer any cancellations to the waitlist.

The clinic scheduling agent sends each patient a confirmation message via their preferred channel — WhatsApp, SMS, email — with a one-tap confirm / reschedule / cancel; the waitlist agent monitors responses in real time; if anyone cancels, the next-best-fit waitlist patient is offered the slot automatically with a 30-minute hold window.

Compliance-aware throughout — no PHI in unsecured channels, no data crossing borders that shouldn't. Audit log of every confirmation and reassignment kept in TRD Storage. No-show rate cut by 40% on average.

compliance-awareauto-reassignaudit log−40% no-shows
Professional services

Onboard a new client without typing the same thing twice

New client just signed — set them up.

The intake agent generates the engagement letter from the firm's template, populated with the client's details; the CRM agent creates the client record; the file-system agent provisions the shared folder; the calendar agent books the kickoff call and a 30-day check-in; the invoicing agent stages the deposit invoice; the welcome agent sends the client a personalized portal link via email and WhatsApp.

Owner reviews the package in one screen — 9 things prepared, all linked, ready to send — approves, and the new client is fully onboarded in under five minutes of human time.

9-step setup5 min human timeone approval
87 of 107 industries

Trained for the business you actually run.

Pilot doesn't act on generic templates — its agents are tuned per industry. 87 industries are live today, on the way to a target of 107. Here's a sample of what's covered, so you can find yours.

87 live·20 in build·107 target
🍽️Restaurants
💇Salons
🏥Clinics
🏠Real estate
🚗Car rental
🛒E-commerce
✈️Travel & tours
⚖️Legal services
🏋️Fitness & gyms
📸Photography
🎓Education
🏨Hospitality
🔧Home services
💼Consulting
🐾Pet services
🎨Creative studios
🏗️Construction
💆Wellness & spa
🚚Logistics
🌱Agriculture
🎤Events
💰Finance
🏦Banking
📊Accounting
🚛Transport
⚙️Manufacturing
🛡️Insurance
📦Shipping
💍Jewelry
👔Fashion retail
🎮Gaming
📱Tech services
🎯Marketing agencies
🍷Bars & lounges
Cafés
+ 52 more industries live

Don't see your industry? Pilot agents are added per vertical — the waitlist form lets you tell us yours, and we prioritize the most-requested ones in the gap from 87 to 107.

3,520 photos · 22 industries · 8 sections each

Every site Pilot ships starts with real photography.

The TRD Image Bank is a prebuilt library of 3,520 photorealistic images — generated by Grok with industry-specific prompts, scored for diversity / hands / faces / composition, and pinned to Filecoin via TRD Storage. Pilot's agents pull from this bank instead of Picsum placeholders, so the first build already feels like the real thing.

22
Industries
8
Sections each
20
Photos per section
3,520
Total photos
Per-industry section structure
hero about services product gallery team menu cta
🍽️
Restaurant
160 photos · 8 sections
✂️
Beauty
160 photos · 8 sections
🩺
Healthcare
160 photos · 8 sections
🏠
Real Estate
160 photos · 8 sections
🛒
Ecommerce
160 photos · 8 sections
💻
SaaS
160 photos · 8 sections
✈️
Travel
160 photos · 8 sections
👗
Fashion
160 photos · 8 sections
🤲
NGO
160 photos · 8 sections
💪
Fitness
160 photos · 8 sections
🛂
Visa
160 photos · 8 sections
⚖️
Law Firm
160 photos · 8 sections
🚗
Car Rental
160 photos · 8 sections
📣
Mkt Agency
160 photos · 8 sections
🎉
Events
160 photos · 8 sections
🏗️
Construction
160 photos · 8 sections
🔗
Crypto
160 photos · 8 sections
💰
Finance
160 photos · 8 sections
📷
Photography
160 photos · 8 sections
🎬
Creator
160 photos · 8 sections
🎨
Portfolio
160 photos · 8 sections
🏪
Amazon
160 photos · 8 sections

Every image is hard-coded with anti-text, hand-binary, anti-collage, and explicit diversity clauses. Each batch is scored PASS / SOFT-FAIL / HARD-FAIL before upload — only PASS images make the bank. Soft-fails go to a regen queue. The result: builds feel real, not stocky.

4,350 templates · 87 industries · 50 each

And 4,350 starting templates behind them.

Every TRD-supported industry ships with 50 production-quality template variations. Different layouts, palettes, content shapes — so two restaurants on TRD don't look like the same restaurant. The orchestrator picks the best-fit template per prompt, then the 641 agents customize from there.

87
Industries live
50
Templates each
4,350
Total templates
107
Industry target
🍽️ Restaurant 50
✂️ Beauty 50
🩺 Healthcare 50
✈️ Travel 50
🛂 Visa 50
🛒 Ecommerce 50
🏠 Real Estate 50
💻 SaaS 50
⚖️ Law Firm 50
💪 Fitness 50
🎓 Education 50
🚗 Car Rental 50
📣 Mkt Agency 50
🎉 Events 50
📷 Photography 50
🏗️ Construction 50
👗 Fashion 50
🤲 NGO 50
💰 Finance 50
🎨 Portfolio 50
🎬 Creator 50
🔗 Crypto / Web3 50
🦷 Dental 50
🛋️ Interior Design 50
📚 Tutoring 50
👕 Laundry 50
🏋️ Sports 50
🎮 Gaming 50
🏪 Marketplace 50
🛒 Grocery 50
🥐 Bakery 50
🧘 Yoga 50
💆 Spa 50
💻 IT Services 50
💍 Jewellery 50
🌸 Florist 50
📊 Consulting 50
🏢 Coworking 50
☀️ Solar 50
🌐 Language School 50
🏨 Hotel 50
🛵 Food Delivery 50
🚘 Car Dealership 50
🧠 Mental Health 50
👔 Recruitment 50
🛡️ Insurance 50
🏛️ Architecture 50
🎵 Music & Artist 50
💒 Wedding 50
🍸 Bar / Nightclub 50
👓 Optometry 50
🩺 Physiotherapy 50
🔒 Cybersecurity 50
📱 App Dev 50
🤖 AI & ML 50
🛋️ Furniture 50
🚗 Driving School 50
🖼️ Art Gallery 50
🔐 Escape Room 50
🎬 Film Production 50
🌍 Translation 50
📦 Moving 50
🐾 Pet Store 50
🕌 Religious Org 50
🌿 Landscaping 50
🔧 Auto Repair 50
🍱 Catering 50
💊 Pharmacy 50
📋 Accounting 50
🎙️ Podcast 50
📰 News & Magazine 50
🧹 Cleaning 50
🔨 Home Services 50
👶 Childcare 50
🎨 Tattoo 50
💡 Life Coach 50
📫 Subscription Box 50
🚚 Logistics 50
🚐 Food Truck 50
🌱 Alt Medicine 50
📚 Bookstore 50
🎤 Comedy 50
🖨️ Print Shop 50
👮 Security 50
🏠 Flooring 50
🕊️ Funeral 50
🗳️ Political 50

Each template is a tested combination of layout + palette + section order + typography — production-quality, not Picsum-grade. Two restaurants on TRD never end up looking like the same restaurant. The orchestrator picks the best-fit template for the prompt; the 641 agents personalize from there.

What makes it different

Six capabilities only TRD Pilot has.

These are not single features — each one is the result of TRD owning a whole layer of the stack. Together they are what makes Pilot able to do what no horizontal agent can.

Vertical multi-agent orchestration

641 industry-specialist agents across 87 live industries (and growing to 107) — restaurants, salons, retail, e-commerce, professional services, fitness, real estate, hospitality, photography, education, construction, and far more. Every horizontal player has one agent trying to do everything. TRD has a fleet of specialists that collaborate.

One product builds + runs the business

TRD Platform generates the site, Compute runs the inference, Storage makes it permanent, OS handles conversations — and Pilot orchestrates them all autonomously. Every competitor solves one slice. TRD is the only closed loop.

Vertical computer use with audit trail

Pilot acts on the platforms you actually use — your point-of-sale, your booking system, your delivery apps, your social accounts. Industry-aware, not generic. Every action emits a signed receipt so you can audit every change.

Predictive outcome engine

Watcher agents observe your bookings, traffic, sales, and competitors. They surface proposed actions before you ask — "your bookings dropped this week, here are three things to try." No competitor does this for small businesses.

Messaging-first interface

The shop owner doesn't open a dashboard at 11pm. You message TRD Pilot on whatever channel you already use — WhatsApp, Telegram, Instagram DM, SMS — and the work gets done. No competitor has a meaningful messaging interface; it doesn't fit their audience.

Small-business pricing

Other autonomous agents cost $200–249/mo. TRD Pilot will start at the same tier as the Platform itself — at small-business prices small businesses have always paid for software. Not a discount: a different cost base because TRD owns the compute layer.

Capabilities in detail

Every feature, grouped honestly.

TRD Pilot ships in five layers, each one building on the last. The roadmap is honest: we tell you which layer ships when, and the status of every feature inside it. Crucially, most of the underlying infrastructure already runs in production today — Pilot is the productization layer that turns the substrate into a single, branded experience.

Already running

The substrate is real. 641 agents are live (verified by SQL), the 5-tier inference chain is in production, agent memory with RAG retrieval already drives WhatsApp replies and build prompts, permanence + audit logs are deployed, the multi-channel inbox aggregates across WhatsApp, Telegram, IG and SMS today, and the Predictive Outcome Engine is live in observability mode. Pilot's job is to wrap all of this into one branded experience the SMB owner can actually use.

Layer 1 · Agents that don't stop.

Phase 1 · Now shipping

Today's AI agents run for a few seconds and stop. You ask a question, it answers, the thread ends. TRD Pilot's foundation is the opposite — agents that take a command and keep going: working for minutes, hours, or until the task is complete, holding memory across sessions, and emitting a signed receipt for every action. This is the layer everything else builds on. Without persistent agents, the rest of the system can't exist.

Long-running mode In progress

Agents that keep working for minutes, hours, or until the task is complete — not just one prompt-response. Give one command in the morning, see the result in the afternoon.

Foundationnon-negotiable

Persistent memory Substrate live

Every agent remembers your voice, your brand, your preferences, your past decisions, your common edits. The underlying layer (agent_memory.ts) shipped April 2026 with RAG retrieval, already feeding the build prompt and WhatsApp replies in production.

Layer · memorysubstrate exists

Signed action receipts Substrate live

Every autonomous action emits a signed receipt — what ran, against which model, for how long, what changed — pinned permanently to TRD Storage. The permanence dashboard (/admin/permanence.html) and audit log (/audit-log.html) are already live; Pilot integrates them under one view.

Layer · trustcross-product

Conversational autonomy Substrate live

Cowork's messaging agents handle inbound conversations autonomously — booking, rescheduling, FAQ, basic upselling — and escalate to you only when a human is genuinely needed. WhatsApp agents are already performing live edits in production; the unified multi-channel inbox is shipped.

Layer · channelCowork-native

MCP server for the 641 agents In progress

Every TRD agent becomes a named MCP tool — restaurant-menu-agent, salon-booking-agent, visa-intake-agent — callable from Claude, ChatGPT, the Coinbase Agent Store, n8n, Make, and Zapier. Inverts io.net's MCP play: the world's AI assistants use TRD's vertical specialists, not the other way around. Internal Tier 0 Inference API already shipped — this is the public MCP-compatible layer (formerly TRD Agents API) on top of it.

Layer · platformMCP-native

Layer 2 · The command center.

Phase 2 · Q4 2026

Once agents persist, the next leap is letting the customer speak to them like a single teammate, not five products. The command center is that interface. One input — on the dashboard or on whatever messaging channel the owner already uses — accepts plain-language commands. The orchestrator parses intent, picks the right agents from the 641 specialists, runs the work, and surfaces a clean diff for one-tap approval. Watcher agents work in the background even when no command is given, flagging what needs attention before the owner notices. The dashboard stops being a CMS and starts being a control center.

Natural-language command bar Planned

One input at the top of the dashboard. "Add a winter promotion banner and tell my regulars about it." The orchestrator parses it, picks agents, runs the workflow, surfaces a diff for one-click approval.

Layer · interface

Always-on watcher agents Substrate live

Background agents that watch your site, your bookings, your inbox, your competitors. They report what's changing and propose actions — you approve, they execute. Retention-email cron and A/B testing observability already run on this pattern; Pilot unifies them.

Layer · monitoringbounded autonomy

Self-healing quality loop Planned

TRD's Quality Scoring system runs on every build. When a score drops, Pilot routes the failing section to a fix agent automatically. Failed contrast → rewritten. Slow load → compressed. Self-correcting sites.

Layer · auto-fix

Reusable workflow templates Substrate live

Sequences you run repeatedly — monthly review, seasonal refresh, new-product launch — become saved one-click templates. Build them once, use them forever. Journey presets are already shipped (Shot 42) — Pilot extends them to whole-business workflows.

Layer · repeatable

Multi-channel coordination Substrate live

One command touches your site, your WhatsApp customers, your Telegram channel, your Instagram, your email list, and your Filecoin archive — all in lockstep, all logged. The unified multi-channel inbox aggregator (WhatsApp + Telegram + IG + SMS) is already shipped; Pilot adds the command-bar that drives it.

Layer · orchestrationclosed loop

WhatsApp x402 agent-pay-agent Researching

The first agent-to-agent commerce protocol with real volume — because TRD's 641 agents are already collaborating across channels. When the restaurant-marketing agent needs a custom social asset, it pays the photo-generation agent in x402 micropayments over WhatsApp. Owner sees one consolidated invoice; agents negotiate, transact, and reconcile autonomously. The agent economy starts here.

Layer · commerceagent-economy

Layer 3 · Acts on the platforms you use.

Phase 3 · Q1 2027

Layers 1 and 2 act on TRD-owned surfaces — your site, your messaging, your asset vault. Layer 3 is when Pilot reaches outside the TRD ecosystem and operates the platforms an SMB owner already uses every day: the point-of-sale, the booking platform, the delivery apps, the marketplace listings, the social accounts, the payment processors. Critically, this is vertical computer use — not the generic kind. The restaurant agent knows the difference between updating a Toast menu and a Zomato listing because it was built for that. Every external action emits a signed receipt so you can audit every change, and the system asks for approval until the owner has built a trust history with each action type.

TRD Vertical Computer Use Planned

The umbrella for Layer 3. Anthropic Computer Use and OpenAI Operator can drive any website — generically. They don't know what makes a Toast menu update different from a Square menu update. TRD does. Per-industry computer-use agents that know Toast / Square / Zomato / OpenTable for restaurants, Booksy / Acuity / Square Appointments for salons, Practo / common EHRs for clinics, Stripe / Razorpay for payments. Generic computer-use is a feature. Vertical computer-use is a product.

Layer · executionvertical-native

Per-industry connectors Planned

Restaurants: Toast, Square, OpenTable, Zomato, delivery apps. Salons: appointment systems, retail POS. Retail: Shopify, marketplaces, inventory tools. The platforms your industry actually uses — not a generic browser.

Layer · integrationvertical depth

Bounded autonomy with audit Planned

Every external action emits a signed receipt. You see exactly what was changed, where, when, and by which agent. Approve-before-action by default; fully autonomous only after a clean track record.

Layer · trust

Multi-account synchronizer Planned

Run a chain with several locations? One agent fans out to all of them simultaneously — same update, every account, in one command. No more copy-paste across five logins.

Layer · scale

Failure-to-human escalation Planned

When the agent can't proceed — a captcha it can't solve, a UI change, an ambiguous call — it pauses, screenshots, and pings you on your preferred channel. You reply; it continues. No silent failures.

Layer · safetynever silent

Layer 4 · Predicts before you ask.

Phase 4 · Q4 2026 · active mode

Layers 1–3 are reactive: you give a command, Pilot executes. Layer 4 turns the system predictive. Watcher agents see your bookings, your traffic, your sales, your competitors. The predictive engine connects the dots and surfaces proposed actions before you notice the underlying signal. "Your Thursday bookings dropped 22% — here are three things competitors with rising bookings did this week. Want me to run any of them?" The system isn't just doing what you ask; it's showing you what to ask — which is the difference between automation and an actual operating system.

Predictive outcome engine In progress

Pilot watches your metrics — bookings, traffic, conversions, competitor moves — and surfaces proposed interventions before you notice the problem. "Your Thursday bookings are tracking 22% below last month. Want to run any of these three things?" Wave Q-1 of the Predictive Outcome Engine is already live in observability mode — activation moves it from watching to acting.

Layer · predictionunique to TRD

Agents that write their own tools Researching

New platform you started using that TRD doesn't know yet? An agent inspects it, drafts a connector, and TRD reviews and ratifies. Long-tail platform coverage at near-zero marginal cost.

Layer · adaptive

TRD Verified badge Researching

Businesses that run TRD Pilot for 90 days with a clean autonomous-action audit log earn a public TRD Verified mark. A real trust signal for your customers — proof that your business is run accountably.

Layer · reputation

Memory-aware industry intelligence (F6) Researching

Cross-customer pattern recognition with privacy preserved. Agents learn that "salons in this city that posted Wednesday promotions saw 18% lift" or "restaurants offering early-bird specials before 7pm cut no-shows 31%." Your data stays yours; the network learns from anonymized patterns and surfaces them back to every member. The flywheel doesn't start until this ships — and only TRD can ship it because only TRD has the 641-agent vertical lens.

Layer · networkflywheel

Multimodal content engine Researching

Text, image, audio, video — from one prompt, in your brand voice. Short-form video for social, voice replies for your inbound calls, multilingual variants in one pass.

Layer · creation

Layer 5 · The vertical operating systems.

Phase 5 · H2 2027

Layer 5 is the destination this whole arc was pointed at. Layers 1–4 are the substrate; Layer 5 packages that substrate into industry-specific operating systems — Restaurant OS, Salon OS, Clinic OS, Retail OS — sold as one product each. A restaurant owner doesn't buy "Platform + Compute + Storage + OS + Pilot"; they buy TRD Restaurant OS and get everything pre-configured for how restaurants actually work. Same shape for salons, retail, and the other major verticals. Beyond TRD's own products, Layer 5 also ships a white-label core for partners who want to launch their own vertical OS, and a framework export for customers who outgrow the hosted product and want to self-host.

TRD Restaurant OS Researching

All TRD products configured for one industry, sold as one product. Site, ordering, menu sync across delivery apps, bookings, loyalty, daily-specials automation, predictive promotions, weekly reporting — one command runs your whole restaurant's online presence.

Vertical OSindustry-first

TRD Salon OS In progress

Same shape, salon-specific. Appointment management, stylist scheduling, no-show recovery, retention campaigns, product upsell, review collection. The "I run my whole salon by texting Pilot" product.

Vertical OS

TRD Clinic OS In progress

Patient intake, appointment management, prescription reminders, compliance-aware comms, common-EHR connectors. Heavier regulatory surface, higher per-customer revenue, narrower initial market — fewer customers, each more valuable. Ships alongside Salon OS in Q3 2026 per the platform roadmap, not in H2 2027.

Vertical OSQ3 2026

TRD Retail OS Researching

For shops and small retail chains. Inventory across channels, marketplace listings, promotion sync, customer loyalty, supplier coordination, sales reporting. One command, all your sales channels in step.

Vertical OS

TRD Business OS (white-label) Researching

The underlying platform that powers all vertical OS products — sold to partners launching their own vertical product for industries TRD doesn't directly target. Partner program, partner marketplace.

PlatformB2B2B

TRD Framework Export Researching

Customers who outgrow the hosted product can export their TRD setup as a self-hosted framework — same orchestrator, same agents, their own infrastructure. Removes the "what if TRD shuts down" objection forever.

Sovereignenterprise grade
Architecture

Why this is hard — and how TRD already solved it.

Building an autonomous-agent product is not just a frontend on top of an LLM API. Five layers of infrastructure have to exist before the customer ever types a command. TRD has been building them for the last 18+ months — the substrate is real, in production, and feeding the products you can already use today.

01 · Orchestrator

Multi-agent routing across 641 specialists

A central orchestrator parses a customer command, decomposes it into sub-tasks, picks the right industry agents for each one, and coordinates their execution — with retry, fallback, and bounded-autonomy guardrails. This is the layer most "AI agent" products skip entirely; it's also the layer that makes vertical specialization possible at scale.

02 · Memory layer

Persistent agent memory across sessions

Every agent retains context across conversations — the customer's voice, brand, preferences, past decisions, common edits. The memory layer is multi-tenant (per-customer isolation), scoped (each agent only sees what it needs), and audit-trailed. You never re-explain yourself.

03 · Inference chain

Five-tier model routing for cost & speed

Different tasks deserve different models. The inference chain routes simple work to fast, cheap models and reserves the heaviest reasoning for tasks that need it — Cerebras for ultra-fast, Together and Gemini for balanced workloads, DeepSeek for cost-sensitive batch. TRD's own Compute network handles increasing portions over time as the community supply scales.

04 · Sentinel

Bounded autonomy & guardrails

Every agent action passes through the Sentinel system before it touches a customer surface. Protected zones, action limits, approval thresholds, sensitive-data scrubbing, and tamper-evident logging are all enforced at the runtime level — not as polite guidance to the model. This is the layer that makes autonomous agents safe to give external account access.

05 · Receipt & storage

Signed audit trail pinned to Filecoin

Every autonomous action emits a signed receipt: which agent ran, against which model, for how long, what changed, on whose behalf. The receipt is pinned to TRD Storage (Filecoin) and surfaced in the customer's dashboard. You can audit any action, undo any change, and prove what happened to anyone who asks — including yourself, six months later.

Security & trust

Trust agents with your accounts? Here's why you can.

Letting an autonomous agent touch your booking system, your delivery apps, or your social accounts is a real psychological step for any business owner. The four guardrails below are how TRD Pilot earns that trust — by default, not as an upsell.

Approval-before-action by default

Every action that touches an external system or sends a customer-facing message requires your one-tap approval before execution. You see a clear summary of what will change, where, and how. Only after a clean 90-day record with a specific action type does the system propose moving it to fully autonomous — and you can revoke that any time.

Per-account scoped credentials

When you connect an external account — your point-of-sale, your delivery platform, your social — Pilot gets a scoped credential that limits exactly what it can do, just for that account. Credentials are stored encrypted, never visible to the agent itself, never logged in plaintext. Revoke a connection in one click and every agent loses access immediately.

Tamper-evident audit log

Every autonomous action emits a signed receipt pinned to TRD Storage on Filecoin. The receipt records what changed, when, by which agent, against which model, for how long, and what it cost. You can replay your entire account history at any granularity — or hand a regulator, an auditor, or yourself six months from now a permanent, mathematical record of every action taken.

Never-silent failure escalation

When an agent can't proceed — a captcha it can't solve, a platform UI change, an ambiguous decision — it pauses, screenshots, and pings you on your preferred channel. You see the exact step, the screenshot, what the agent was trying to do, and you reply with the right call. The agent picks up and continues. No silent failures, no surprise consequences.

Region-aware compliance. TRD operates across five regions in parallel — US, Europe, MENA, Asia, and ANZ — and Pilot inherits the platform's region-aware data handling: GDPR for European customers, HIPAA-aware patterns for healthcare verticals, data-residency controls per region. Sensitive data is scrubbed before it reaches any model that doesn't strictly need it.
Roadmap

Named, dated, honest.

Five phases over 18 months. We tell you what ships when, and we don't pretend something is ready when it isn't.

Phase 1 · Shipping now (was Q3 2026)

Agents that don't stop — substrate already live

The foundation is real today. Memory layer (agent_memory.ts) with RAG retrieval is in production, the 5-tier inference chain is live, the permanence dashboard and audit log are deployed, and WhatsApp agents are already performing autonomous live edits. What ships as "TRD Pilot" is the productization of this substrate — branded, packaged, and exposed to customers as one experience.

Memory liveReceipts liveCowork liveAPI in progress
Phase 2 · Q3–Q4 2026

The command center — UI on top of running substrate

Natural-language command bar, always-on watcher agents, self-healing quality loop, reusable workflow templates. Most pieces already exist (journey presets, retention cron, A/B testing, multi-channel inbox); Phase 2 is the unified front-end that lets the SMB owner drive everything from one input.

Command bar newWatcher liveTemplates live
Phase 3 · Q1 2027

Acts on the platforms you use

The biggest genuinely new layer: vertical computer use across industry-specific platforms — point-of-sale, booking systems, delivery apps, marketplaces, social accounts, payments. Bounded autonomy, multi-account fan-out, never-silent failure escalation.

Computer useConnectorsSynchronizer
Phase 4 · Q4 2026 · active mode (was Q2 2027)

Predicts before you ask — engine already exists

Wave Q-1 of the Predictive Outcome Engine is already deployed in observability mode (Zone 75) — watching metrics, accumulating training signal. Phase 4 is the activation step: moving from observation to action, exposing the suggestions to the owner, and integrating with the command bar. Multimodal content engine and Verified Badge ship alongside.

Engine liveActivation Q4Multimodal
Phase 5 · Q3 2026 (Salon/Clinic) → H2 2027 (rest)

The vertical operating systems — earlier than expected

Salon OS and Clinic OS are already on the Q3 2026 platform roadmap as full vertical template sets — they ship alongside Pilot Phase 1, not in H2 2027. Restaurant OS and Retail OS follow through 2027. White-label TRD Business OS and Framework Export close out the year.

Salon OS Q3Clinic OS Q3Restaurant OSWhite-label
The TRD ecosystem

Pilot is the layer that runs the rest.

Most autonomous-agent products are one product. TRD Pilot orchestrates four others — each a real product, each handling a different layer of running a business — into one autonomous system.

01 · Platform

Builds the site

The AI website builder. One prompt turns into a finished multi-page site for your business.

02 · Compute

Powers the work

The community GPU network that runs the inference behind every action — owned demand, not speculative.

03 · Storage

Makes it permanent

Filecoin-backed permanence for everything Pilot does — your site, your assets, every signed receipt.

04 · OS

Handles conversations

Conversational Commerce OS for WhatsApp, Telegram, Instagram, SMS — every customer message, autonomously.

05 · Pilot

Runs them all

The orchestration layer. Tell Pilot what you want; it picks agents from the right products and does the work.

Why this matters. A horizontal agent product can drive a browser. TRD Pilot orchestrates a website builder, a GPU network, a permanence layer, a messaging platform, and 641 industry agents. That combination doesn't exist anywhere else — and it's the only combination that lets one command run your whole business end to end.
Global from launch

Built for five regions in parallel.

Every other autonomous-agent product on the market today is US-only, US-first, or enterprise-only with no SMB regional path. TRD Pilot launches into five major regions simultaneously — because small businesses with one to fifty employees exist in identical density across every developed and emerging market.

United States

33M+ small businesses

Stripe-native payments, English-first. The most competitive SMB market and the most underserved by vertical agents today.

Europe · EU + UK

GDPR-compliant

Multi-language: German, French, Spanish, Italian, Dutch, Polish. Strong SMB density, strong messaging-app penetration.

MENA

Mobile-first, messaging-dominant

Arabic RTL support, WhatsApp-native usage patterns. Fast-growing SMB segment across UAE, Saudi Arabia, Egypt, Morocco.

Asia · South + SE

Messaging-first markets

India, Indonesia, Vietnam, Philippines, Thailand. Multi-language at launch — Hindi, Bahasa, Vietnamese, Thai, Tagalog. Local payment rails supported.

Australia + NZ

Mature SMB market

English-native, Stripe-native, low-friction onboarding. High per-customer lifetime value. Strong fit for the upper tier.

Pricing

Three tiers. Small-business prices.

Final numbers will be confirmed closer to Phase 1 launch in Q3 2026. The shape below is locked: three tiers, billed monthly, the same structure as TRD's existing pricing — not the $200–$249/month enterprise tier the horizontal autonomous-agent products live in. Waitlist members get the introductory pricing.

Pilot Starter
$ indicative
Solo owners, single location
  • Layers 1–2: persistent agents + command center
  • Single business / single location
  • One messaging channel — WhatsApp or Telegram or SMS
  • Up to 100 autonomous actions per month
  • 30-day audit history retention
  • Email support, 48h response
Join waitlist →
Pilot Pro
$ indicative
Multi-location chains, professional firms, agencies
  • Everything in Growth, plus —
  • Layer 4: predictive outcome engine + watcher agents
  • Multi-account synchronizer for chains
  • Up to 10,000 autonomous actions per month
  • Multi-user team accounts with per-role permissions
  • Custom industry agent training on your own data
  • Dedicated success manager, 2h response
Join waitlist →
How this works honestly. TRD owns the underlying Compute network and most of the inference happens on the community GPU layer — which is why TRD Pilot can be priced where small-business software has always lived, not at the $200/month tier the horizontal autonomous-agent products charge. It's not a discount; it's a different cost base. Layer 5 (Restaurant OS, Salon OS, Clinic OS, Retail OS) ships as standalone products with their own pricing later in 2027 — you'll be able to bundle or buy them separately.
Honest scoping

What TRD Pilot won't do.

We've watched too many AI products oversell what their agents can actually do — and then customers feel misled. So here is the honest list of things Pilot is not built for, and won't pretend to handle.

Not for this

Replacing human relationships with your customers

Pilot can handle reservations, confirmations, FAQ, and routine messages autonomously. But the actual customer relationship — the loyal regular, the difficult complaint, the personal touch — is yours. We deliberately make Pilot easy to silence on any thread the moment you want to take over.

Not for this

Replacing your accountant, your lawyer, or your doctor

Pilot is not a regulated professional. It does not give legal, tax, medical, or financial advice. It can prepare invoices and route financial data; it can schedule appointments and message patients in compliance-aware ways; it can draft a document. But it does not substitute for the professionals your business actually needs.

Not for this

Anti-detect browsing or platform-policy evasion

Pilot operates real accounts via real, sanctioned interfaces — official APIs where they exist, browser automation where they don't, but always within each platform's terms of service. We do not build tooling for fingerprint spoofing or anti-detection. That's a different industry, and not one we want to be in.

Not for this

Enterprise IT integration projects

Pilot is built for small and mid-sized business owners who want to type one message and have things happen. It is not built for enterprise integration projects requiring six-month rollouts, an IT department, and a steering committee. That's what Microsoft Copilot Studio and ServiceNow Project Arc are for — they own that market, we don't try to.

Not for this

Fully autonomous actions on day one

By default every action requires your approval. Fully autonomous mode is unlocked per action type, only after a clean 90-day record, and is always revocable. We will not pretend agents can be trusted from minute one — they can't, and any product that says otherwise is going to disappoint a customer.

Not for this

Replacing tools you already love

If your accounting tool works, your CRM works, your delivery platform works — Pilot connects to them, it doesn't replace them. The Vertical OS products (Restaurant OS, Salon OS, etc.) are the integrated path for people who want the all-in-one product. Everyone else uses Pilot as the orchestration layer above the tools they already trust.

FAQ

Honest answers, now.

TRD Pilot is in active development. Here is exactly what's ready, what isn't, and what to expect.

What is TRD Pilot, in one sentence?

TRD Pilot is an autonomous AI layer that runs your business across every tool you use — your site, your customer messaging, your booking system, your delivery apps, your social accounts — from a single command in plain language.

How is this different from ChatGPT, Copilot, or Operator?

Two structural differences. First, those products are horizontal — one agent that tries to do everything. TRD Pilot is vertical — 641 specialist agents across 87 live industries (on the way to 107) that already know how restaurants, salons, retail, professional services, fitness, real estate, hospitality, photography, construction, and 80 others actually work.

Second, they're priced and built for power users, developers, or enterprise IT departments. TRD Pilot is built and priced for the small-business owner running a shop, a salon, a clinic, or a studio.

When can I actually use it?

Layer 1 (persistent agents) ships in Q3 2026. Each subsequent layer ships every three to six months after, with the vertical operating systems shipping on a staggered schedule: Salon OS and Clinic OS in Q3 2026, Restaurant OS and Retail OS through 2027.

The waitlist gets early access at each layer's launch, plus the introductory pricing tier.

How much will it cost?

Pricing for TRD Pilot will be announced closer to Q3 2026, but the principle is locked: it sits in the same range as TRD's existing tiers (small-business prices, not enterprise prices). For comparison, the major autonomous-agent products today start at $200–$249/month and are aimed at developers or enterprises.

What kinds of businesses is it for?

Restaurants and food service. Salons and beauty. Retail shops and small chains. Fitness studios and gyms. Clinics and small healthcare practices. Real estate brokers. Tour operators and travel agencies. Professional services firms. E-commerce founders. Crafts and creators with their own storefronts.

If your business is between one and fifty employees and your customers come through a mix of website, messaging, and walk-in, TRD Pilot is built for you.

Which regions does it work in?

Global from launch — United States, Europe (EU + UK), MENA, Asia (South + Southeast), Australia and New Zealand. Unlike most autonomous-agent products today, TRD Pilot is not US-only and not US-first.

Does it work in my language?

TRD's underlying platform supports 10 languages today, expanding to 18+ in Q4 2026 including English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Arabic (with RTL), Hindi, Bahasa, Vietnamese, Thai, Tagalog, Japanese, Korean, and Chinese. Pilot inherits the same language support.

What if I don't trust an agent touching my accounts?

By default, every action requires your approval before execution. You see a clear summary of what will change, where, and how, and you click to approve. Only after you've built a clean 90-day record with a particular type of action do we propose moving it to fully autonomous.

Every action also emits a signed receipt pinned to TRD Storage — a permanent, tamper-evident record you can audit any time.

Can I try it before official launch?

Yes — waitlist members get early access at each layer's launch, plus discounted introductory pricing. Members are also invited to private beta programs for specific industry verticals before the public Restaurant / Salon / Clinic / Retail OS products ship.

What if I'm already using TRD Platform, Compute, Storage, or OS?

Pilot is built specifically to orchestrate those products. If you're already on the TRD Platform, your site, your messaging, your asset library, and your audit history all become things Pilot can act on from day one — no integration work required. Existing TRD customers will get an upgrade path and an introductory credit on Pilot when Phase 1 ships.

Does Pilot work if I'm not on the TRD Platform yet?

Yes — Pilot will connect to external sites and external platforms in Phase 3 regardless of whether you build with TRD Platform. The experience is richest when all your products are in the TRD stack (Platform + OS + Storage + Pilot working together), but Pilot does not require it. You can use Pilot as the orchestration layer above any existing tools you already trust.

What if an agent does something I didn't want?

Every autonomous action is reversible. One tap rolls back the change, and the signed receipt of the original action remains in your audit log so you can see exactly what happened. For actions that touch external platforms, Pilot also keeps a "before" snapshot of the state it changed, so reverting is exact, not approximate.

For action types where you've enabled fully autonomous mode and want to undo it, you can revoke that permission and return to approval-before-action with one click.

How are the 641 agents actually trained?

Each industry agent has been built since April 2024 through a mix of prompt engineering, retrieval against industry-specific reference material, and tested against real customer builds. Agents are tuned per industry — the restaurant agent set is not the salon agent set is not the visa-consultant agent set. The whole fleet runs through TRD's orchestrator and shares the same memory and audit infrastructure.

Pilot Pro customers will be able to fine-tune specific agents on their own data for their own business — a salon chain can teach its agents the brand voice, the specific terminology, the local nuances.

What happens if TRD itself goes down or shuts down?

Two answers, depending on the layer. Your data is already addressed: TRD Storage mirrors everything you've built to Filecoin, so even if TRD's servers disappeared overnight, your site, your assets, and your audit trail remain retrievable through any public IPFS gateway.

Your operations are addressed by Phase 5's Framework Export. Pro customers can export their TRD setup — orchestrator, agents, configuration — as a self-hosted framework they run on their own infrastructure. The "what if you shut down" objection is one TRD has built around from day one.

How does this compare to AdsPower, Multilogin, or other RPA tools?

Those tools record clicks and replay them — RPA is reactive script playback. It breaks the moment a website's UI changes, and it doesn't reason about what to do. Pilot's agents reason: they understand the goal, pick the steps, adapt when something changes, and emit a signed receipt of what was done.

If your work is multi-account browser automation (the AdsPower use case), those tools may still be the right pick. If your work is running a business, Pilot is a different category of product.

Will my employees still have jobs?

Different question for different team sizes, but the honest answer for the small businesses Pilot is built for: the work Pilot takes off is mostly the work owners already do at 11pm because no one else has time — menu updates across delivery apps, social posts, customer confirmations, weekly reporting, repetitive coordination tasks. The human-judgment work — service quality, relationships, strategy, complex decisions — is exactly what Pilot makes more space for.

For multi-person teams, Pilot's approval flow and audit trail are designed so team members keep oversight and ownership of the work they care about, while getting the busywork off their plates.

Is TRD a real company or a research project?

Real company. DCS AI Technologies L.L.C is the legal entity. TRD Platform, Compute, Storage, and OS are all live products. Pilot is the next-generation autonomous layer being added to that stack — built on infrastructure that has been running in production for over a year. The waitlist is real, the roadmap is real, and the team ships on schedule.

Can investors, partners, or press get more detail?

Yes — for investor inquiries, strategic partnerships, white-label or framework-export discussions, or press, please reach out via the contact links in the footer. We share more detailed technical, financial, and roadmap material under NDA where appropriate.

Be first when TRD Pilot goes live.

One email when Phase 1 launches in Q3 2026. Plus early access, discounted introductory pricing, and an invite to the industry-specific private betas. No drip campaign, no sales pitches.

~3,000+ already on the TRD waitlists across products.