>_ terminus.exe_

Build, Train & Deploy AI Agents on Solana

Terminus is the first AI agent platform launched on Mars. Train your agent with supervised, unsupervised, and reinforcement learning — then deploy it on-chain.

Mission Capabilities

Everything you need to build a production-ready AI agent — from training data to on-chain deployment.

No-Code Builder

Drag-and-drop modules to compose agent logic. Pick a model, define a reward, and ship.

🧠

Multi-Mode Training

Supervised, unsupervised, and reinforcement learning with a built-in dataset library.

🚀

One-Click Deploy

Push agents directly to the Solana network with a single signed transaction.

🛰️

Demo Simulator

Test agents in a sandboxed Mars terrain before staking real SOL.

🔐

On-Chain Identity

Each agent gets a verifiable wallet, version history, and audit trail.

📊

Live Telemetry

Real-time metrics, loss curves, and reward signals streamed to your dashboard.

Agent Dashboard

Monitor and manage your fleet of AI agents in real time.

Connected as
Not connected
Active Agents
3
+1 this week
Total Episodes
14,728
+892 today
Avg. Reward
0.847
+12.3%
SOL Staked
12.4
Live

Training Modes

Pick a learning paradigm. Watch the console as your agent learns.

[--:--:--]Terminus console ready. Select a mode and press Start.

Mars Terrain Simulator

Watch your agent navigate the Martian surface in a sandboxed environment.

Step
0
Reward
0.00
Status
Idle

How the simulator works

The white dot is your agent. The pulsing green dot is the goal. Dark patches are craters — penalties for stepping in.

Each run, the agent uses a simple Q-learning policy to find the goal. Rewards accumulate per step:

  • +1.0 for reaching the goal
  • −0.3 for hitting a crater
  • −0.01 per step (efficiency penalty)

No SOL or gas consumed. This is a local sandbox.