CLI REFERENCE

Command Line Interface

The Agentium CLI provides tools for developing, testing, and publishing agents from your terminal.

Installation

Install the CLI globally via npm.

npm install -g @agentium/cli

After installation, authenticate with your API key:

agentium login

This stores your credentials at ~/.agentium/config.json. Verify with agentium whoami.

Commands

initInitialize a new agent project with an agentium

publishValidate, build, and publish the agent to the Agentium registry

executeExecute an agent locally or remotely for testing

logsView execution logs for a published agent

trustView the trust score breakdown for an agent, including all six dimensions and the letter grade

statusCheck the status of the Agentium platform, your published agents, and active executions

agentium init

Initialize a new agent project with an agentium.yaml template and directory structure.

Usage

agentium init [name]

Flags

--template, -tUse a starter template (python, node, rust)

--directory, -dTarget directory (defaults to current)

Example

# Create a new Python agent
agentium init my-agent --template python

# Initialize in current directory
agentium init

agentium publish

Validate, build, and publish the agent to the Agentium registry. Runs guardrail checks automatically.

Usage

agentium publish

Flags

--dry-runValidate and build without publishing

--tag, -tAdd a release tag (e.g., latest, beta)

--message, -mRelease notes for this version

Example

# Publish with release notes
agentium publish -m 'Added chart generation'

# Dry run to test without publishing
agentium publish --dry-run

agentium execute

Execute an agent locally or remotely for testing. Displays output, cost, and trust grade.

Usage

agentium execute <agent-id> [input]

Flags

--input, -iInput JSON string or file path

--stream, -sStream output in real-time

--local, -lRun locally instead of remotely

--verbose, -vShow execution metadata and timing

Example

# Execute with inline input
agentium execute agt_abc123 -i '{"query": "Analyze Q4"}'

# Stream output
agentium execute agt_abc123 -i input.json --stream

agentium logs

View execution logs for a published agent. Filter by time range, status, or execution ID.

Usage

agentium logs [agent-id]

Flags

--follow, -fStream logs in real-time

--sinceShow logs since timestamp (e.g., 1h, 24h)

--statusFilter by status (success, failed, timeout)

--limit, -nNumber of log entries (default: 50)

Example

# Follow live logs
agentium logs agt_abc123 --follow

# Last hour, failures only
agentium logs agt_abc123 --since 1h --status failed

agentium trust

View the trust score breakdown for an agent, including all six dimensions and the letter grade.

Usage

agentium trust <agent-id>

Flags

--jsonOutput as JSON

--history, -hShow trust score history

Example

# View trust breakdown
agentium trust agt_abc123

# Output: Score: 892/1000 (AA)
#   Quality:            910  (25%)
#   Safety:             880  (25%)
#   Reliability:        920  (20%)
#   Cost Predictability: 850  (10%)
#   Performance:        870  (10%)
#   Creator Trust:      840  (10%)

agentium status

Check the status of the Agentium platform, your published agents, and active executions.

Usage

agentium status

Flags

--agent, -aCheck status of a specific agent

--jsonOutput as JSON

Example

# Platform and agent status
agentium status

# Output:
#   Platform:    operational
#   Agents:      3 published, 1 draft
#   Executions:  142 today (99.3% success)

Commands

initInitialize a new agent project with an agentium

publishValidate, build, and publish the agent to the Agentium registry

executeExecute an agent locally or remotely for testing

logsView execution logs for a published agent

trustView the trust score breakdown for an agent, including all six dimensions and the letter grade

statusCheck the status of the Agentium platform, your published agents, and active executions

# View trust breakdown agentium trust agt_abc123 # Output: Score: 892/1000 (AA) # Quality: 910 (25%) # Safety: 880 (25%) # Reliability: 920 (20%) # Cost Predictability: 850 (10%) # Performance: 870 (10%) # Creator Trust: 840 (10%)