Spectra

Landing page
Bot Live Analytics
Manual Agent Configuration
Real Time API Tracking
Kill Switch
Discord Live Updates Of Bot Anamoly
Tech Stack Mermaid Diagram

Inspiration

The statistics for AI agents on Solana are exploding, with frameworks allowing autonomous agents to manage their own treasuries and trade millions in volume. But there is a glaring, terrifying gap: Who is watching the agents?

Every developer's nightmare is an agent "hallucinating" or suffering a logic loop and draining money in seconds. We noticed that while the tools to build agents are fast, the infrastructure to secure them is non-existent. Traditional monitoring is built for humans, not for high-frequency autonomous bots.

We asked ourselves: What if you had a "safety camera" for the autonomous economy? What if you could catch a rogue agent before it spent it was too late? What if you could see the exact moment a trading bot deviated from its baseline behavior?

That’s when Spectra was born.

What it does

Spectra is a real-time behavioral classification and analytics platform that acts as a security layer for AI agents on the Solana blockchain.

Core Features

REAL-TIME TRANSACTION SURVEILLANCE

- Direct RPC Integration: Monitors agent wallets via Solana RPC onLogs listeners for instant visibility.

- Transaction Parsing: Automatically decodes complex Solana instructions into human-readable trades, swaps, and transfers.

- Live Stream: A dedicated activity feed that updates in real-time as the agent interacts with the network.

ANOMALY DETECTION ENGINE

- Frequency Analysis: Detects "Spam Attacks" or logic loops by monitoring transaction density.

- Velocity Tracking: Calculates spending velocity over rolling windows to prevent treasury drains.

- Alerting System: Triggers visual and system alerts the moment an agent breaches its pre-set safety parameters.

"ELECTRIC" GLASSMORPHISM DASHBOARD

- High-Fidelity UI: A cutting-edge "Dark Electric" aesthetic designed for security command centers.

- Ambient WebGL Layer: A background of WebGL threads representing network connectivity and neural activity.

- Mobile Optimized: Fully responsive frosted-glass components that look professional on any device.

BEHAVIORAL CLASSIFICATION

- Persona Categorization: Automatically labels agent behavior as Conservative, Aggressive, or Suspicious.

- Baseline Comparison: Compares current agent activity against historical norms to identify subtle deviations.

ROGUE SIMULATION TOOLKIT

- Inject Anomaly: A built-in developer tool to force an agent into "Rogue Mode" for testing oversight response.

- Deterministic Testing: Simulate "Bad Trades" or rapid-fire transactions to verify that security thresholds are working.

How we built it🛠️

┌─────────────────────────────────────────────────────────┐
│                 Frontend (Next.js 14)                   │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐ │
│  │WebGL     │  │Shadcn/UI │  │TypeScript│  │Tailwind  │ │
│  │Threads   │  │Components│  │          │  │CSS       │ │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘ │
└─────────────────────────────────────────────────────────┘
                              │
                              ▼
┌─────────────────────────────────────────────────────────┐
│               API Layer (Next.js API Routes)            │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐ │
│  │Agent     │  │Anomaly   │  │Solana    │  │Simulation│ │
│  │Management│  │Detector  │  │Watcher   │  │Engine    │ │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘ │
└─────────────────────────────────────────────────────────┘
                              │
                    ┌─────────┴─────────┐
                    ▼                   ▼
┌──────────────────────┐    ┌──────────────────────┐
│   Blockchain Layer   │    │   Database Layer     │
│  ┌──────────────────┐│    │  ┌────────────────┐  │
│  │Solana Devnet     ││    │  │Supabase        │  │
│  │- Web3.js         ││    │  │- Agent Config  │  │
│  │- RPC Listening   ││    │  │- Tx History    │  │
│  │- Log Parsing     ││    │  │- Anomaly Logs  │  │
│  └──────────────────┘│    │  └────────────────┘  │
└──────────────────────┘    └──────────────────────┘

- Frontend: Built with Next.js 14 and React for high-performance dashboarding, using WebGL Threads to visualize network energy.

- Solana Pipeline: Leverages @solana/web3.js and onLogs RPC subscriptions to monitor the Solana Devnet with sub-second latency.

- Anomaly Detection: A specialized engine that calculates transaction frequency f and spending velocity V to identify "Rogue" agent states.

- Data Persistence: Supabase (PostgreSQL)

- Simulation Engine: A custom-built testing suite that allows developers to "Inject Anomaly" to verify the oversight logic under pressure.

Challenges we ran into

The biggest headache was definitely the Git workflow. We ran into a nightmare of "divergent branches" while trying to push our pricing and support pages, which forced us to learn the hard way how to reconcile local work with remote updates without losing code.

Accomplishments that we're proud of

We are hyped that our "Rogue Mode" trigger actually works. Seeing the dashboard instantly turn red and blast out alerts when we force an agent to "go bad" is incredibly satisfying. We also managed to build a UI that doesn't just look like a "hackathon project", it looks like a professional security product you'd see in a real startup.

What we learned

Building on Solana taught us that Trust Engineering is the next big frontier. Speed is great, but oversight is what makes that speed safe for enterprises. We leveled up our skills in WebGL, real-time RPC handling, and complex Git management.