Dialectic Self-Debate Arena
Four autonomous agents debate architecture, critique edge cases, and synthesize solutions.
The Proposer
Thesis FormulationThe Adversary
Critique & VulnerabilitiesThe Researcher
Fact & Reference VerificationThe Judge
Synthesized ConsensusGoal Loop Pipeline
Iterative task formulation, adversarial critique, and hardened output generation.
Stage 1: Formulation
Initial solution draft and code synthesis
Stage 2: Adversarial Review & Score
Flaw detection, edge-case probing, and quality rating
Stage 3: Hardened Delivery
Refined, production-ready output
🕸️ Agent Swarm Canvas
Consensus: ReadyAutonomous Multi-Agent Directed Acyclic Graph (DAG) orchestration: Architect ➔ Systems Engineer ➔ Threat Auditor ➔ Adversarial QA ➔ Chief Synthesizer.
🎯 Master Swarm Deliverables
Thinking Profile Studio
Manage user preferences, coding styles, and active RAG notes across PC and Mobile in ~/.antri/profiles/.
profile_1.md
Editing thinking profile instructionsMarkdown (.md) Skill Studio
Specialist capability directives loaded from ~/.antri/skills/. Add or import new skills as Markdown files.
Select a Skill
Click any skill on the left to inspect or edit instructionsLifelong Memory & Vector Store
128-dimensional dense semantic embeddings, episodic interaction logs, and workspace conventions.
Semantic Memory Store
Episodic Interaction Logs
💡 Suggestions & Creative Ideation Studio
Explore 2-3 tailored architectural blueprints, design themes, and feature ideas. Select any idea or provide your own customized requirements!
🎨 Custom Architecture & Personal Vision
Adaptive CustomizerHave your own idea or want to modify one of the suggestions above? Type your thoughts below and the agent will build it with maximum craftsmanship.
🎨 Artifacts Hub & Interactive Studio
Browse and interact with generated BugTwin sandboxes, CrashZero time-travel replays, mind maps, and live SPA artifacts.
🫁 Codebase Intelligence Radar
Pre-analyzed codebase cache, technology stack matrix, subsystem modules, and Git workspace status.
💻 Live Code Workspace & Playground
Browse workspace files, edit source code directly, and preview live HTML artifacts in real-time sandbox.
🧠 Autonomous Engineering Mission Control
Deep multi-stage thinking pipeline, architectural planning protocols, and anti-shallow code verification.
🫁 Codebase Breathing & Intelligence Cache Retrieval
On boot, ANTRI takes 1-2 seconds to index manifests, technology stack, directory layout, and entrypoints into active memory cache without blocking.
🏛️ Deep Architectural Blueprint & Type Schema Design
Before coding, ANTRI models domain entities, DTOs, data flow, responsive breakpoints, state management, and edge cases to eliminate 4-5 line shallow codebases.
🧱 Modular File Scaffolding & Multi-File Generation
Writes real physical source files directly to the project directory (index.html, style.css, app.js, package.json, components) with zero fake placeholders.
🛡️ Defensive Engineering & High-Craft Polish
Integrates Web Audio API sound synthesizers, keyframe animations, Chart.js visualizations, Canvas-Confetti celebrations, and comprehensive error handling.
🌐 Dual-Delivery Synergy
Real production files safely persisted on disk for deployment & Git, plus instant interactive single-page application artifact for in-chat desktop preview.
⚙️ Cloud Run & Multi-Key Settings
Configure Google Cloud Run backend endpoints, multi-API key auto-rotation pools, and platform sync.
☁️ Google Cloud Run Backend (.run.app)
Local NodeConnect ANTRI Desktop & CLI to your Google Cloud Run container. Automatically inherits Gemini 3.7 / 3.5 models and tools.
🔑 Multi-API Key Auto-Rotation Pool
1 Key ActiveProvide 3-4 API keys. When one key reaches rate limits (429) or runs out of credits, ANTRI seamlessly spins up the next key with zero downtime.