AAgentis Docs

Quick Start

Create an Agentis account, install the Agentis skill, and let your AI agent operate within policy.

Agentis is built for humans who want AI agents to hold and spend money without handing them raw wallet access.

The intended path is:

  1. Sign in to the Agentis dashboard.
  2. Create an agent wallet.
  3. Fund it and set policy limits.
  4. Install the Agentis skill in your AI agent environment.
  5. Ask your agent to use Agentis for paid fetches, transfers, policy checks, privacy, or yield.

The dashboard is the fastest place to start. The CLI, SDK, MCP server, and skill let you keep controlling the same agent wallets from terminals, apps, and AI agent environments.

1. Sign in

Open the Agentis dashboard and sign in.

Use the dashboard to get the first wallet and policy in place:

  • create hosted agent wallets
  • inspect balances and transaction history
  • fund an agent wallet from your connected wallet
  • set spend limits and kill switches
  • enable privacy flows for private agents
  • review Jupiter Earn positions

2. Create an agent wallet

Create a new hosted agent from the dashboard.

Give it a name you will also use when talking to your AI agent, for example research-agent, buyer-agent, or dev-agent.

Agentis creates a Solana wallet for that agent. You can fund it, cap what it is allowed to spend, and turn it off from the dashboard at any time.

3. Set policy before handoff

Set limits before your AI agent starts spending.

Common first policies:

  • a max amount per transaction
  • a daily budget
  • a monthly or total budget
  • an allowed-domain list for paid API calls
  • a kill switch you can turn on instantly

Policy amounts are USD-denominated. Agentis checks policy before signing or proxying payments.

4. Install the Agentis skill

Install the Agentis skill in the AI agent environment you use for coding or automation:

npx skills add awesamarth/agentis

The skill teaches your agent when to use the dashboard, CLI, SDK, MCP server, policy checks, Umbra privacy, Jupiter Earn, and paid MPP/x402 fetches.

After installing it, tell your agent what you want it to do:

Log me into Agentis and list my wallets.
Fetch this paid endpoint via one of my wallets with enough balance.
Put all my agents' available USDC balances into Jupiter Earn.
Put 0.05 SOL from dev-wallet into Umbra encrypted balance.

5. Install the CLI

For most coding agents with shell access, the Agentis CLI is the best execution path.

Install the CLI once, then your agent can use the agentis command directly. The skill gives it the context it needs to discover commands with agentis --help, pick the right interface, and operate wallets without you memorizing syntax.

Install:

npm install -g @agentis-hq/cli

Then ask your AI agent to use Agentis. It can log you in, list wallets, check balances, select a funded wallet, run paid fetches, update policies, use Umbra, or deposit into Jupiter Earn through the CLI.

Use the CLI docs when you want to see the full command set yourself.

6. Use MCP only when you do not want the CLI

MCP is the secondary path for AI agent environments where you do not want a global CLI installed, or where the host already has Agentis MCP tools connected.

Use the MCP docs when you want a local assistant to call Agentis tools through stdio instead of shelling out to agentis.

Use the SDK docs when you are building Agentis into an app backend or custom agent runtime.

What your agent can do

Once the agent wallet, policy, skill, and interface are in place, your AI agent can use Agentis to:

  • check balances and transaction history
  • pay MPP and x402 protected URLs
  • send SOL through a policy-checked wallet
  • update or read spend policy
  • use Umbra privacy actions for private agents
  • inspect or deposit into Jupiter Earn
  • scaffold or publish x402 facilitators

You keep control through Agentis. Your agent gets a constrained financial tool it can actually use.

Manual paths

The CLI, SDK, and MCP server are still first-class Agentis surfaces, but they are not the best first step for most users.

Use them when you need direct control:

  • CLI: terminal workflows, scripts, local encrypted wallets, and manual operations
  • SDK: app backends and custom agent runtimes
  • MCP: local AI assistants and coding agents that should call Agentis tools directly

For day one, create the agent wallet, install the skill, connect an Agentis interface, and let your agent operate with policy in place.