Open Console →
● How it works

From "idea" to runs by itself, forever

Four simple steps turn an AI into a safe, self-running agent. Then we'll show you the real breakthrough underneath — the part that finally lets things run on a blockchain without any bot or server you own.

Jump to the breakthrough ↓New here? Start with the basics

The agent lifecycle

Four steps — one stays in your hands

1

🚀 Boot

One action creates the agent and gives it everything at once: a readable name (mybot.rama), its own wallet, a sponsored-gas budget, and a fresh rulebook. No accounts to set up, no keys to copy around.

One transaction · fully on-chain
2

🛡️ Constrain

You set the limits: how much it can spend per day, who it's allowed to pay, which actions need your approval, and how long its working key stays valid. These rules are enforced by code — the agent literally cannot break them.

You hold the master key · it gets a limited "session key"
3

⏰ Schedule

Tell it what to do and when — "rebalance my savings every Monday", "pay this every 1st". The task is stored on the blockchain with a small prepaid budget. This is the part that used to be impossible without a server. (More below 👇)

Runs on time · no cron, no server
4

👁️ Observe & control

Every action is written to a public ledger you can read. Approve big moves, change the limits, pause the agent, or revoke it completely — instantly, from anywhere.

Human always in control
🚀 Bootname · wallet · gas 🛡️ Constrainlimits · approvals ⏰ Scheduleon-chain · prepaid 👁️ Observepause · revoke you stay in control at every step

The breakthrough

Smart contracts were powerful — but asleep

A blockchain smart contract can hold a rule like "pay rent on the 1st of every month." But here's the catch almost nobody talks about: a contract cannot run itself. Blockchains only do something when someone sends them a transaction. The contract just sits there, waiting to be poked.

😴 The old way — a hidden weak spot

Every "automatic" crypto app secretly relied on an off-chain bot / cron server running somewhere to wake the contract up.

  • Single point of failure. If that one server goes down, your "automation" silently stops.
  • Not really decentralized. The whole thing depends on one operator's machine and wallet.
  • Fragile & opaque. Runs out of gas money, gets misconfigured, or the owner disappears — and it just… stops.

⚡ The Ramestta way — the schedule lives on-chain

The task, its timing, and the money to run it are all stored on the blockchain. An open market of independent keepers competes to run due tasks and earn the fee.

  • No server you own. If one keeper is offline, another runs it — because they're paid to.
  • Prepaid on-chain. The task carries its own budget; execution is permissionless and self-funded.
  • Truly autonomous. Automation with no hidden off-switch and no single operator to trust.
🤖 Agentregisters a task ⏰ On-chain Scheduler"run at 09:00 daily" + prepaid fee ⚙️ Keeper A — earns the fee ✓ ⚙️ Keeper B — standby ⚙️ Keeper C — standby many keepers compete → the task always runs, on time

The agent is the brain (decides what & when). The on-chain scheduler + keeper market is the heartbeat (makes sure it actually happens).

🧠
AI decides

The agent works out what to do and when — the smart, flexible part.

⛓️
Blockchain guarantees

The scheduler + rulebook make sure it happens on time and inside your limits — the reliable, trustless part.

💸
Keepers deliver

An open, paid market of executors keeps it running with no server for you to maintain.

This is the piece that was missing — self-running smart contracts, without a bot in the middle.

The safety model

The agent never holds the key that owns everything

This one idea is what makes autonomy safe.

🗝️
Master key = you

It stays in your wallet. It's the only key that can change the rules or take everything back.

🎟️
Session key = the agent

A temporary, limited key the runtime uses to act. It expires, and you can revoke it anytime — without moving a single coin.

Approvals for big moves

Anything above a limit you set pauses for a human "yes". Small routine actions just flow.

See it in real life

What can you actually build with this?