> ## Documentation Index
> Fetch the complete documentation index at: https://docs.glam.systems/llms.txt
> Use this file to discover all available pages before exploring further.

# AI

You can manage GLAM vaults through AI coding agents using the [glam-skill](https://skills.sh/glamsystems/glam-skill/glam). Instead of running CLI commands or clicking through the GUI, describe what you want in natural language and the agent handles the rest.

The skill gives your agent a decision tree for vault operations, error handling, integration awareness, and knowledge of the GLAM CLI and SDK.

### Supported platforms

The glam-skill works with any [skills.sh](https://skills.sh)-compatible platform:

* Claude Code
* Gemini CLI
* Codex
* OpenCode

### Installation

For skills.sh-compatible platforms:

```bash theme={null}
npx skills add https://github.com/glamsystems/glam-skill --skill glam
```

<Info>
  In Claude Code, you can also install with `/skill` or add the skill manually to your project configuration.
</Info>

### Configuration

Before your agent can manage vaults, you need the GLAM CLI installed and configured.

Install the CLI:

```bash theme={null}
npm install -g @glamsystems/glam-cli
```

Create your CLI config at `~/.config/glam/config.json`:

```json config.json theme={null}
{
  "keypair_path": "~/.config/solana/id.json",
  "json_rpc_url": "https://api.mainnet-beta.solana.com",
  "priority_fee": { "micro_lamports": 5000 },
  "jupiter_api_key": "<jupiter-api-key>"
}
```

<Info>
  Need an RPC endpoint? AI agents can create a [Helius](https://www.helius.dev/docs/agents/overview) account and get an API key automatically. Your `json_rpc_url` would then be `https://mainnet.helius-rpc.com/?api-key=<helius-api-key>`.
</Info>

<Info>
  The `jupiter_api_key` is optional but recommended. It's required for commands that fetch token prices (e.g., `vault token-balances`). Get one for free on [Jupiter Developer Platform](https://portal.jup.ag/pricing).
</Info>

The agent uses this config automatically when running CLI commands. If you have multiple keypairs (e.g., owner, delegate, user), you can either update the config or tell the agent to use a specific config file with the `-C` flag.

### Example walkthrough

Here's what managing a GLAM vault looks like through an AI agent. The following prompts mirror the [CLI quickstart](/v1/quickstart/cli) flow.

<Steps>
  <Step title="Create a vault">
    Prompt your agent:

    ```
    Create a tokenized GLAM vault called "AI Quickstart USDC" with USDC as
    the base asset. Set max cap to 1000 USDC, min subscription to 1 USDC,
    and zero fees.
    ```

    The agent will generate the vault configuration JSON and run the appropriate `glam-cli vault create` command.
  </Step>

  <Step title="Enable integrations">
    ```
    Enable Kamino Vaults integration on my vault.
    ```

    The agent knows which integrations are available and runs:

    ```bash theme={null}
    glam-cli integration enable KaminoVaults
    ```
  </Step>

  <Step title="Grant delegate permissions">
    ```
    Grant my delegate ndcXmT...ids4QnG permissions to fulfill mint requests
    and deposit/withdraw from Kamino Vaults.
    ```

    The agent translates this into the correct `delegate grant` commands with the right protocol and permission names.
  </Step>

  <Step title="Manage portfolio">
    ```
    Deposit 5 USDC into Kamino lending (main market).
    ```

    The skill knows the main Kamino market address, so this works out of the box. If you want to use a different market or vault, just paste the address in your prompt.
  </Step>

  <Step title="Handle investor flows">
    ```
    Fulfill all pending redemption requests.
    ```

    The agent runs `glam-cli manage fulfill` to process queued redemptions.
  </Step>
</Steps>

### What the skill provides

* **Decision tree** for vault operations -- the agent knows which commands to run and in what order
* **Error handling** -- the agent interprets CLI errors and suggests fixes
* **Integration awareness** -- the agent understands which integrations need to be enabled before use and what permissions delegates need
* **CLI and SDK coverage** -- the agent can use either the CLI or the TypeScript SDK depending on your context

### Learn more

See the full skill reference at [skills.sh/glamsystems/glam-skill/glam](https://skills.sh/glamsystems/glam-skill/glam).