01
Durable operational memory
Local neural memory, conversations, projects, plans, approvals, checkpoints, research ledgers, schedules, and results persist across restarts.
AGI Model 10 reasoning · durable operational memory · real tools
The AI that does not
stop at an answer.
Thaddeus places approved business calls, sends approved texts, works across 1,000+ apps, supervises active jobs by live voice, runs durable automation, drives your real computer and browser, builds verified software, and keeps the project trail on your machine.
Apple Silicon Mac available now. Windows x64 opens after publisher signing.
1,000+supported app integrations
100+actions across major ad platforms
Local-firstdurable projects and memory
Real callsoptional managed Calling Agents
The operating layer
One workspace. The whole job.
Every base plan includes the core operating system below. Thaddeus can reason, remember, act, verify, and resume—not simply produce another chat response.
02
Computer Copilot
See and operate supported Mac and Windows workflows across browsers, files, terminals, Office apps, Xcode, and other visible tools—with permission controls.
03
1,000+ apps that can act
Connected accounts become executable tools: read, create, update, send, publish, analyze, and manage supported resources—not a decorative connector list.
04
Advertising operations
Work with Meta Ads, Google Ads, and LinkedIn Ads to create or update campaigns, creatives, audiences, budgets, and reporting when the required credentials and permissions are connected.
05
App Builder
Build polished, responsive web apps, stores, portals, dashboards, tools, and games with isolated project files, real tests, a working preview, visual review, and publish-ready delivery.
06
Native software workflows
Use coding tools and Computer Copilot with Xcode for native iOS work and release steps. Apple credentials, signing, spending, and final App Store submission remain under your approval.
07
Business Suite
Create bounded AI employees, manage customers and call history, prepare offers and campaigns, run routines, and track externally verified results from one business workspace.
08
Calling Agents
Provision a managed business number, configure goals, consent, transfer rules, hours, voices, and tests, then launch approved calls and review their outcomes.
09
Deep Research with honest results
Inspect exact source excerpts, run an adversarial counter-evidence lane, and see every claim graded Supported, Contested, Inference, Uncertain, or Unsupported before exporting PDF or Markdown.
10
Dreamify studio
Create and edit images, animate stills, generate or extend video, work from references, and produce voice inside the same account with visible usage.
11
Artifacts + deliverables
Produce presentations, reports, proposals, guides, infographics, ads, social assets, and interactive deliverables with durable revisions and portable exports.
12
Relay + scheduled work
Pair a phone by one-time QR to send jobs, watch progress, answer questions, approve actions, and stop work—or run durable one-off and recurring jobs on a schedule.
13
Thaddeus Reactor
Keep a live voice command lane open while terminal, computer, or background work continues. Ask for status, queue another job, stop active work by voice, or wake it with “Hey Thaddeus.”
14
Android texting
Pair a compatible Android phone through an isolated Google Messages session and send approved SMS from the user's real number. The paired phone must remain powered on and online.
15
Cooperative browser
Work inside a separate persistent browser profile where sign-ins stay on the computer, hand control to a person when needed, gate consequential actions, and record or replay bounded workflows.
16
Plan-first execution
Turn work into a structured objective, steps, verification, evidence, and blockers; approve an exact plan revision before execution and compare the finished delivery with that contract.
17
Multi-agent workforce
Launch bounded background agents and coordinated workflows, follow their status, move eligible work to the background, and collect durable deliverables without freezing the foreground conversation.
18
Real shell + bundled runtime
Run actual project commands through guarded Bash or PowerShell lanes with bundled runtime and search tooling, native notifications, deep links, signed updates, and visible stop controls.
19
Local-first project state
Conversations, memory, plans, app projects, checkpoints, research inputs, schedules, browser workflows, and generated media are stored locally with bounded files and hardened persistence.
20
Unity bridge · private preview
A first-party Unity editor bridge is under active private testing. It is not included in every public build yet and will be advertised as generally available only after release verification.
Meta AdsGoogle AdsLinkedIn AdsGmailGoogle CalendarSlackHubSpotSalesforceShopifyStripeGitHubNotion+ hundreds more
Connection truth: available actions, authorization method, scopes, rate limits, and account eligibility vary by platform. Thaddeus asks for the access a job needs and keeps consequential external actions behind its approval boundaries.
From intent to outcome
Give it the operation, not a sequence of prompts.
Thaddeus combines its own workspaces instead of making you move fragments between separate chat, browser, automation, media, and business products.
Launch and operate a campaign
Move from research and creative to connected ad execution and verified results.
- Research audience, offer, and competitors
- Create images, video, copy, and landing assets
- Build the campaign page or working web app
- Operate supported ad accounts with approval
- Inspect performance and preserve the result
Build and ship software
Keep requirements, code, tests, screenshots, revisions, and release work in one durable trail.
- Create a complete, tested web product in App Builder
- Review a real localhost journey and visual evidence
- Revise from a saved checkpoint without losing working features
- Use coding tools and Xcode for native iOS work
- Pause before publishing, credentials, or paid actions
Run a customer operation
Turn a business objective into repeatable work that can continue from desktop or phone.
- Define an AI employee and its approval limits
- Manage customers, offers, campaigns, and routines
- Launch a disclosed Calling Agent from a managed number
- Schedule recurring jobs and keep run history
- Approve or stop work remotely through Relay
Why Thaddeus
Memory is useful. Operational continuity is different.
ChatGPT and Claude both offer cross-chat memory. Thaddeus goes further by keeping the actual operation durable: project files, app revisions, research evidence, plans, schedules, approvals, run history, and connected-tool state can be reopened and continued.
BUILT TO OPERATE
It can actually call for you.
Launch disclosed outbound AI calls to approved domestic-rate +1 destinations from a managed number. Calling Agents use prepaid minute packs and remain under your control.
ChatGPTNo native managed outbound Calling Agent
ClaudeNo native managed outbound Calling Agent
ThaddeusManaged number · agent setup · call records
Swipe the table to compare →
| Capability | ChatGPTGeneral assistant | ClaudeKnowledge work | ThaddeusOperations agentACTION + CONTINUITY |
|---|---|---|---|
| Memory | Cross-chat memory for personal context, preferences, files, and connected-app context when enabled | Memory summaries across chats and projects on eligible plans | Local neural memory plus durable operational state for projects, builds, evidence, schedules, workflows, approvals, and results |
| Computer action | Cloud-browser and app-based agent capabilities; availability varies | Cowork can use Mac and Windows desktop apps through Claude Desktop when the computer is online | Native Computer Copilot for supported Mac and Windows desktop workflows plus an isolated working browser and explicit trust audit |
| Connected-app execution | Apps can search, reference, sync, and take supported actions; availability varies | Connectors and MCP tools can expose supported actions; availability varies | 1,000+ supported app integrations exposed as executable tools inside the same autonomous workflow |
| Ad operations | Possible through available apps, browser workflows, or custom tooling | Possible through connectors, Cowork, or custom tooling | Named Meta Ads, Google Ads, and LinkedIn Ads toolkits with 100+ catalog actions across campaigns, creative, audiences, budgets, and analytics |
| Software delivery | General coding and agent tools | General coding, artifacts, and Cowork workflows | Dedicated App Builder with isolated files, deterministic quality checks, localhost journey, screenshot-bound visual review, revisions, and publish-ready packaging |
| Business operations | General assistant and app workflows | General knowledge-work and connector workflows | Dedicated Businesses workspace for AI employees, customers, Calling Agents, offers, campaigns, routines, calls, and verified results |
| Live voice supervision | Voice conversation; no cited first-party lane for supervising a separate active desktop job | Voice conversation; no cited first-party lane for supervising a separate active desktop job | Reactor remains conversational while work runs: ask status, queue work, stop work, and receive spoken completion notices |
| Durable automation | Scheduled one-off and recurring tasks, monitoring, and notifications; supported tools and limits vary | Scheduled Cowork tasks can use connected tools, skills, and plugins, run remotely, and preserve past runs | Local cron schedules, one-off runs, reusable job templates, approvals, foreground/background lanes, retry/crash recovery state, and durable run history execute through a dedicated worker |
| Phone-to-desktop control | Tasks and agent surfaces can be reached across supported devices | Mobile Dispatch can assign and monitor Cowork or coding tasks and reach desktop apps while the desktop is available | Relay pairs directly to this desktop by one-time QR with a per-device revocable credential for jobs, live progress, answers, approvals, stops, and results |
| Android SMS | No native paired Android SMS sender in the cited plans | No native paired Android SMS sender in the cited plans | Pair Google Messages in an isolated persistent session and send approved SMS from the user's Android number |
| Browser continuity | Agent browsing uses a managed cloud browser for supported public sites | Cowork and extensions can work with web content and desktop surfaces | Separate persistent local browser profile, human sign-in handoff, action approvals, and recorded/replayable workflows with duplicate-action protection |
| Plan-to-delivery contract | Planning and agent execution are available | Planning and agent execution are available | Approve an exact structured plan revision, then track evidence and the finished delivery against its objective, verification, and blockers |
| Creation suite | Image generation and voice features | File and image analysis, artifacts, and voice features | Dreamify image, image editing, video generation/editing/extension, references, and voice plus a separate Artifact Studio |
| Project data model | Cloud service with account controls and some desktop/local features | Cloud service with projects, Cowork, and desktop/local features | Local-first file-backed conversations, memory, projects, checkpoints, media, browser workflows, research inputs, schedules, and delivery history |
| Outbound AI calling assistant | No native managed agent | No native managed agent | Yes Managed number, readiness setup, consent controls, voices, and prepaid minute packs |
| Usage visibility | Limits vary by plan and feature | Rolling five-hour limits plus weekly limits on paid plans | Published Compute Units, five-hour windows, and an in-app meter |
Comparison methodology and official sources
Checked July 19, 2026. Competitors do have memory, scheduled work, connected tools, and agent capabilities; this table distinguishes those from Thaddeus's combined local operating stack and dedicated verification gates. Sources: ChatGPT memory, ChatGPT Work cloud browser, ChatGPT apps, ChatGPT scheduled tasks, ChatGPT voice, ChatGPT Business, Claude project memory, Claude plans, Claude connectors, Claude Cowork, Claude scheduled tasks, Claude mobile Dispatch, and Claude voice. Thaddeus integration action counts reflect the live upstream catalog on the check date: Meta Ads, Google Ads, and LinkedIn Ads. Features vary by plan, platform, region, account eligibility, settings, and permissions. Competitor names are trademarks of their respective owners.
Base plans
More usable work. Clear limits.
Compute Units track actual workload—not raw message counts. Efficient tasks use less capacity, while deep reasoning, long output, media, and extended workflows use more. Your in-app meter keeps usage visible as you work.
Starter
$99/month
20M Compute Units
5M per five-hour window
- All core desktop features
- AGI Model 10 high reasoning
- Dreamify access
- Voice and app connections
MOST POPULAR
Pro
$199/month
50M Compute Units
15M per five-hour window
- All core desktop features
- Larger daily workflows
- More media capacity
- Priority for serious users
Elite
$499/month
125M Compute Units
40M per five-hour window
- All core desktop features
- Heavy autonomous operation
- Large research/build runs
- Highest monthly capacity
Founder
$4,999/year
750M Compute Units
200M per five-hour window
- Annual capacity
- Founder account status
- Roadmap feedback channel
- Limited availability
How resets work: base capacity renews on your billing date. A five-hour window starts with your first metered request after the previous window expires. Unused capacity does not roll over. Provider outages, abuse controls, concurrency, and safety limits can still pause work.
Secure checkout is being finalized. No charge was made.
Optional add-ons
Scale the capabilities you use most.
Start with a complete Thaddeus plan, then add dedicated Calling, Dreamify, or Compute capacity as your workload grows. Every add-on stays flexible under the same account, so you can expand one capability without changing your entire plan.
Calling Agents
Outbound calling to approved domestic-rate +1 destinations, one managed number, one active call, and up to ten minutes per call. Voice transport comes from the pack; the agent's model responses use base Compute Units.
Dreamify capacity
Dedicated monthly media units used before your base balance, giving frequent image and video creation its own visible capacity.
$49/mo+10M media unitsAdd Boost
$149/mo+30M media unitsAdd Studio
General Compute
Extra monthly Compute Units for chat, reasoning, tools, Calling intelligence, images, and video when your workload grows.
Add-ons require an active Thaddeus base plan, renew monthly, expire without rollover, and may be canceled independently.
Verified installers only
Download for your platform.
macOS · Apple Silicon
Signed and notarized public release. Install from the DMG into Applications.
Download .dmg →
Windows 11 · x64
Private runtime QA is complete. Public download remains closed until Delphi Labs Inc. publisher signing is verified.
Publisher signing in progress
Subscribe securely
Stripe collects payment, phone, billing details, and affirmative acceptance of the Terms.
Install the verified build
Use only this site. Do not bypass an unknown-developer or unknown-publisher warning.
Sign in with checkout details
Enter the exact Stripe email and the code delivered to the checkout mobile number.
FAQ
Straight answers.
Does Thaddeus only connect apps, or can it use them?
It can execute the actions exposed by an authorized integration—such as creating, updating, sending, publishing, or analyzing supported resources. The exact actions depend on the app, your account eligibility, credentials, granted scopes, and platform limits.
What makes its memory different?
ChatGPT and Claude have cross-chat memory. Thaddeus also stores file-backed operational state: projects, app files and revisions, research evidence, plans, schedules, approvals, workflows, and results can survive a restart and be continued instead of reconstructed from a summary.
What is Thaddeus Reactor?
Reactor is the live voice command layer. It can keep talking while another eligible job runs, report foreground and background status, queue a new work instruction, stop active work, and announce when background work succeeds, fails, or needs attention.
Can it send real text messages?
With a compatible Android phone paired to its isolated Google Messages session, Thaddeus can send approved SMS from that phone's real number. The phone must be powered on and online. Inbox reading, inbound synchronization, and MMS are not advertised as public features today.
Does local-first mean no data ever leaves the computer?
No. Core project state and generated files are stored locally, but requested model, media, telephony, web, and connected-app operations may transmit the data needed to the relevant service. Permissions, the Privacy Policy, and each connected service still apply.
Can it manage advertising?
Yes, after you connect eligible accounts and required credentials. The current integration catalog includes Meta Ads, Google Ads, and LinkedIn Ads actions for campaigns, ads and creative, audiences, budgets, targeting, and analytics. Thaddeus still follows approval and platform-permission boundaries.
Can it build and launch an iOS app?
Thaddeus can write and test native code and operate supported Xcode workflows through Computer Copilot. Its dedicated App Builder currently produces web apps and publish-ready web packages. Apple signing, developer-account access, paid actions, and final App Store submission require your approval.
How much actual work does a plan provide?
There is no useful fixed message count: a quick answer uses far less capacity than a long autonomous build, deep reasoning run, or video. The in-app meter shows your live balance as you work.
Why a five-hour window?
It lets normal work resume automatically after a break while preventing one runaway job or compromised key from consuming an entire month at once.
Is Dreamify included?
Access is included and can use base Compute Units. Optional Dreamify packs keep a separate media-only balance for customers who create heavily.
Why are Calling Agents separate?
Calling combines live phone-network and speech transport for every connected minute. Dedicated packs make minute capacity predictable and keep base plans focused on the work every customer uses.
Can Calling Agents contact anyone?
No. Launch is outbound-only to valid +1 destinations, and you must already have every legally required consent, honor do-not-call and opt-out rules, and disclose the AI and your organization.
Can I bring my own model key?
Not in the public plans. Provider credentials remain server-side so routing, privacy boundaries, exact usage, and abuse controls stay consistent.
Can I cancel?
Yes. Cancel before renewal through the billing portal. Access continues to the end of the paid period unless suspended for abuse or legal risk. Charges and unused capacity are normally non-refundable.