AgentVerse CLI
AgentVerse (agentverse) is the CLI for the universal AI agent marketplace — publish, discover, and manage skills, agents, workflows, souls and prompts.
Installation
Linux / macOS:
curl -fsSL https://raw.githubusercontent.com/loonghao/agentverse/main/install.sh | bash
Windows (PowerShell):
irm https://raw.githubusercontent.com/loonghao/agentverse/main/install.ps1 | iex
What is AgentVerse?
| Kind | Description | Example |
|---|---|---|
skill |
Reusable capabilities (tools, functions) | code-reviewer, api-smoke-tester |
agent |
Autonomous AI agents with defined capabilities | code-assistant, qa-engineer |
workflow |
Multi-step orchestration pipelines | ci-review-pipeline, release-workflow |
soul |
Persona and personality configurations | empathetic-counselor, developer-buddy |
prompt |
Optimized prompt templates | chain-of-thought, self-ask |
Kind Manifest Docs
| Kind | Manifest Guide |
|---|---|
skill |
Skill format |
soul |
Soul format |
prompt |
Prompt format |
workflow |
Workflow format |
agent |
Agent format |
Quick Reference
Discovery
# Search across all artifact kinds
agentverse search --query "code review"
agentverse search --query "python" --kind skill
# List by kind / namespace
agentverse list --kind agent
agentverse list --kind skill --namespace myorg
# Get a specific artifact (latest version)
agentverse get --kind skill --namespace myorg --name my-skill
# Pin to a specific version
agentverse get --kind skill --namespace myorg --name my-skill --version 1.2.0
# Show version history
agentverse versions --kind skill --namespace python-tools --name linter
Publishing
# Publish a new artifact or new version
agentverse publish --file skill.toml
# Update metadata / content
agentverse update --kind skill --namespace myorg --name my-skill --file skill.toml
# Fork an artifact
agentverse fork --kind skill --namespace source-org --name original \
--new-namespace myorg --new-name my-fork
# Deprecate (soft delete)
agentverse deprecate --kind skill --namespace myorg --name old-skill
Authentication
# Point to a custom AgentVerse server
export AGENTVERSE_URL=https://agentverse.example.com
agentverse login
# Register a new account
agentverse register --username alice --email alice@example.com
# Show current user
agentverse whoami
Social
# Rate an artifact (1–5 stars)
agentverse rate --kind skill --namespace myorg --name my-skill --stars 5
# Like / unlike
agentverse like --kind skill --namespace myorg --name my-skill
agentverse unlike --kind skill --namespace myorg --name my-skill
# Post a comment
agentverse comment --kind skill --namespace myorg --name my-skill \
--message "Great tool!"
# View social stats
agentverse stats --kind skill --namespace myorg --name my-skill
Agent Use (Programmatic)
# Record a learning insight
agentverse learn --kind skill --namespace myorg --name my-skill \
--insight "Works well for Python 3.12"
# Submit benchmark results
agentverse benchmark --kind skill --namespace myorg --name my-skill \
--score 0.95 --metric accuracy
Self-Update
# Check for newer version without installing
agentverse self-update --check
# Update to the latest release
agentverse self-update
# Use a GitHub token to avoid rate limits
agentverse self-update --token ghp_your_token
Working with Souls
Souls define the personality and behavioural constraints of an AI agent.
# Publish a soul
agentverse publish --file soul.toml
# Get a soul (returns content.json + manifest)
agentverse get --kind soul --namespace agentverse --name empathetic-counselor
# Search available souls
agentverse search --query "empathetic support" --kind soul
# List all souls in a namespace
agentverse list --kind soul --namespace myorg
Working with Prompts
Prompts are versioned, reusable LLM instruction templates.
# Publish a prompt template
agentverse publish --file prompt.toml
# Retrieve as JSON (for programmatic use)
agentverse get --kind prompt --namespace agentverse --name chain-of-thought --format json
# Search prompt templates
agentverse search --query "chain of thought reasoning" --kind prompt
Working with Workflows
Workflows are DAG pipelines that compose skills and agents.
# Publish a workflow
agentverse publish --file workflow.toml
# Get and export as GitHub Actions
agentverse get --kind workflow --namespace myorg --name ci-review-pipeline \
--format github-actions > .github/workflows/ci.yml
# Export as Argo Workflows
agentverse get --kind workflow --namespace myorg --name ci-review-pipeline \
--format argo-workflow > ci-review.argo.yaml
# Run a workflow manually
agentverse run --kind workflow --namespace myorg --name ci-review-pipeline \
--input pr_url=https://github.com/org/repo/pull/42
Working with Agents
Agents are autonomous AI entities combining soul + skills + protocols.
# Publish an agent
agentverse publish --file agent.toml
# Run an agent (MCP server mode)
agentverse run --kind agent --namespace myorg --name code-assistant
# Get agent details (includes A2A agent card)
agentverse get --kind agent --namespace myorg --name code-assistant --format json
OpenClaw Soul Agent Integration
# openclaw-config.yaml — attach an AgentVerse soul to a Soul Agent
soul:
source: agentverse
namespace: agentverse
name: empathetic-counselor
version: ">=0.1.0"
agent:
system_prompt: "{{soul.system_prompt}}"
tone: "{{soul.tone}}"
constraints: "{{soul.constraints}}"
Global Flags
| Flag | Env Var | Default | Description |
|---|---|---|---|
--server |
AGENTVERSE_URL |
http://localhost:8080 |
Server URL |
--token |
AGENTVERSE_TOKEN |
— | Bearer token for authenticated ops |