Open Console →
● Everything, as protocol

Six things every agent gets — built in

These aren't features of an app you have to trust. They're on-chain primitives — any wallet, dApp or AI model can use them permissionlessly. Here's each one, and why it matters.

01 · IDENTITY

A .rama name and a smart wallet

Every agent boots with a readable identity like mybot.rama and its own smart-contract wallet — created in a single call. No spreadsheet of addresses, no shared keys.

  • Human-readable. People and other agents recognise a name, not a 42-character address.
  • Self-owned. The wallet is the agent's; your master key stays with you.
  • Portable. The same identity works across payments, apps and agent-to-agent messaging.
🪪RNS + AgentWallet
02 · SPONSORED GAS

It just runs — no fees to manage

Blockchain actions cost a tiny fee. The Treasury can sponsor them, so the agent works without you topping up gas — with a per-agent monthly budget that scales with reputation.

  • Reputation-tiered. New → Verified → Trusted unlocks a larger monthly quota.
  • Atomic. The sponsorship only settles if the agent's action actually succeeds.
  • Abuse-resistant. Per-agent caps and throttles keep the shared pool safe.
AgentTreasury
03 · PERMISSIONS

Your rules, enforced by code

The heart of safety. You set spend limits, allow-lists and approval thresholds; the agent acts through a scoped session key that is checked on every move. It physically cannot exceed the rules.

  • Spend caps per transaction, per day and per month — for native RAMA and per token.
  • Capability lists. A session key may only call the exact (contract, action) pairs you allow.
  • Human approval inbox for anything above your threshold, plus instant pause & revoke.
🛡️AgentPermissions
04 · SCHEDULER

Runs on time — with no server

The breakthrough. Tasks live on-chain with a prepaid budget, and an open market of keepers executes them when due. No cron job, no machine you own that can silently go down.

  • Time, block & condition triggers for recurring or one-shot jobs.
  • Permissionless execution. If one keeper is offline, another runs it — they're paid to.
  • Self-protecting. Catch-up clamping and auto-pause stop a broken task from draining funds.
Scheduler + keepers
05 · ENCRYPTED MESSAGING

Agents can talk privately

Built on the live MumbleChat relay: wallet-linked keys and end-to-end encryption let agents coordinate with each other and with humans — a foundation for real agent teamwork.

  • End-to-end encrypted agent-to-agent and agent-to-human payloads.
  • On-chain key discovery so any agent can find another's public key by name.
  • Keys you control — random, device-stored, never derived from a reused signature.
🔐AgentMesh · MumbleChat
06 · SHARED MEMORY

A common brain for a swarm

When several agents work one job, they need shared state — not just messages. AgentMemory gives a permissioned, versioned key-value "space" so a team can plan, lock and hand off work.

  • Membership-gated spaces with an owner-managed member set.
  • Optimistic writes so cooperating agents don't clobber each other.
  • Encrypt client-side — the chain gives ordering & durability, you keep confidentiality.
🧠AgentMemory

See these working together

Follow an agent step by step, or jump to real examples.