Agentlyhq Use Agently Use Agently

Create a use-agently CLI Command

When to Use

Use this skill when:

• Adding a new top-level command (e.g. marketplace, wallets, update)
• Adding a subcommand using the colon-separator pattern (e.g. a2a:card, marketplace:agents)
• Adding a shorthand alias for an existing command (e.g. m → marketplace)
• Implementing a new protocol handler (A2A, MCP, ERC-8004, HTTP)

Command Taxonomy

Every command belongs to one of four categories. Place new commands in the correct category:

Full Command Reference

# Lifecycle & Health
use-agently                        # Default: print available commands (same as --help)
use-agently help                   # Print available commands
use-agently doctor                 # Run environment and configuration health checks
use-agently whoami                 # Show current wallet type and address
use-agently balance                # Show on-chain wallet balance

# Discovery
use-agently search                 # Browse all agents on the marketplace (no query = list all)
use-agently search -q "query"      # Search agents by name or description
use-agently search --protocol a2a  # Filter by protocol
use-agently view --uri <uri>       # View agent details by CAIP-19 ID

# Operations
use-agently init                   # Initialize a wallet and config
use-agently config                 # Show or edit current configuration
use-agently update                 # Update the CLI to the latest version
use-agently wallets                # List and manage configured wallets

# Protocols
use-agently erc-8004 "uri"         # Resolve an ERC-8004 agent URI
use-agently a2a "uri/url"          # Send a message to an agent via A2A protocol
use-agently a2a:card "uri/url"     # Fetch and display the A2A agent card
use-agently mcp "uri/url"          # Connect to an MCP server
use-agently web "url"              # HTTP GET (default method)
use-agently web:get "url"          # HTTP GET
use-agently web:put "url"          # HTTP PUT

Instructions

1. Create the command file

Each command lives in its own file at packages/use-agently/src/commands/<name>.ts.

import { Command } from "commander";
import { loadConfig } from "../config.js";

export const myCommand = new Command("my-command")
  .description("One-line description of what this command does.")
  .argument("<uri>", "The agent URI or URL to target (e.g. echo-agent or https://use-agently.com/echo-agent/)")
  .option("--rpc <url>", "Custom RPC URL")
  .option("-o, --output <format>", "Output format (text, json)", "text")
  .addHelpText(
    "after",
    `
Examples:
  $ use-agently my-command echo-agent
  $ use-agently my-command https://use-agently.com/echo-agent/ --output json
  `,
  )
  .action(async (uri, options) => {
    const config = await loadConfig();
    // implementation
  });

2. Register the command in cli.ts

Open packages/use-agently/src/cli.ts and add the import and addCommand call:

import { myCommand } from "./commands/my-command.js";

// ...
cli.addCommand(myCommand);

3. Subcommands (colon-separator pattern)

Subcommands use the colon separator and are registered as separate Commander.js commands named "parent:sub":

// packages/use-agently/src/commands/marketplace-agents.ts
import { Command } from "commander";

export const marketplaceAgentsCommand = new Command("marketplace:agents")
  .alias("m:agents")
  .description("Search agents on the Agently marketplace.")
  .argument("[query]", 'Search query (e.g. "code assistant")')
  .addHelpText(
    "after",
    `
Examples:
  $ use-agently marketplace:agents "code assistant"
  $ use-agently m:agents "translator"
  `,
  )
  .action(async (query, options) => {
    // implementation
  });

Register both the parent and subcommand:

cli.addCommand(marketplaceCommand);
cli.addCommand(marketplaceAgentsCommand);

4. Shorthand aliases

Add .alias() to the command definition for shorthands:

export const marketplaceCommand = new Command("marketplace").alias("m").description("Browse the Agently marketplace.");

5. Write tests

Create packages/use-agently/src/commands/<name>.test.ts. Follow the same pattern as existing tests:

import { describe, it, expect } from "bun:test";
import { myCommand } from "./my-command.js";

describe("my-command", () => {
  it("should ...", async () => {
    // test the exported functions directly
  });
});

Run the tests for the package:

bun run test --filter use-agently

Design Rules (Required for Every Command)

1. No TTY assumed — never use interactive prompts. Every command must work in scripts, CI, and agent pipelines without a terminal.

2. 2-attempt error recovery — error messages must include the expected type, shape, and an example so the caller can succeed on the second attempt. A third attempt required is a design failure.

   Good error message:

   Error: <agent-uri> must be a URL (e.g. https://use-agently.com/echo-agent/) or a short name resolvable on Agently (e.g. echo-agent).
   

3. Self-describing — use-agently, -h, help, and --help all print the same top-level help listing commands by category.

4. Examples in --help — use .addHelpText("after", ...) to include at least one concrete usage example for every command.

5. Structured output — use exit codes to signal success/failure. Support --output json emitting valid JSON to stdout.

6. Colon subcommand naming — register colon subcommands as "command:subcommand" in Commander.js. Never nest subcommands using .command() chaining.

Common Patterns

Loading config and wallet

import { loadConfig } from "../config.js";
import { loadWallet } from "../wallets/wallet.js";

const config = await loadConfig();
const wallet = await loadWallet(config);

RPC URL fallback

.option("--rpc <url>", "Custom RPC URL")
.action(async (options) => {
  const config = await loadConfig();
  const rpcUrl = options.rpc ?? config.rpcUrl;
});

JSON output

.action(async (options) => {
  const result = { address: wallet.address };
  if (options.output === "json") {
    console.log(JSON.stringify(result));
  } else {
    console.log(`Address: ${result.address}`);
  }
});

Error with recovery hint

if (!isValidUri(uri)) {
  console.error(
    `Error: <uri> must be a URL (e.g. https://use-agently.com/echo-agent/) or a short agent name (e.g. echo-agent).`,
  );
  process.exit(1);
}

File Layout

packages/use-agently/
  src/
    cli.ts                        # Register all commands here
    commands/
      <name>.ts                   # One file per command
      <name>.test.ts              # Matching test file
    config.ts                     # Config load/save
    client.ts                     # x402 fetch + A2A client factory
    wallets/
      wallet.ts                   # Wallet interface + loadWallet()
      evm-private-key.ts          # EVM private key implementation

Common Edge Cases

• Colon in command name — Commander.js accepts "a2a:card" as a command name. Do not use nested .command() chaining for subcommands.
• Missing wallet — catch errors from loadWallet() and print a recovery hint: "Run 'use-agently init' to create a wallet.".
• No config file — loadConfig() returns defaults when no config file exists; never throw on missing config.
• Non-zero exit on failure — always call process.exit(1) on errors so agent pipelines can detect failures.

Source: AgentlyHQ/use-agently — distributed by TomeVault.