Skip to main content

agentvisor template

Manage agent project templates — list, preview, and scaffold new projects.

Synopsis

agentvisor template <subcommand> [options]

Subcommands

SubcommandDescription
listShow available templates with descriptions
showDisplay the full README for a template
createScaffold a new project from a template

list

Show all available templates with difficulty level, tags, and description.

agentvisor template list

A template is identified by a single <framework>/<name> token, shown in the TEMPLATE column. That token is exactly what show and create accept, so it can be pasted straight from this output. The framework prefix is always part of the name — several templates are called research-agent (LangGraph, CrewAI, ADK, Strands), and only the prefix tells them apart.

Options

OptionDefaultDescription
--framework-Only show templates for this framework
--tag-Only show templates with this tag

Example Output

On a TTY, list renders a styled, color-coded table sized to the terminal. Piped or redirected (as below), it falls back to plain aligned text wrapped at 80 columns — the same data, grep-friendly, with long descriptions truncated:

TEMPLATE LEVEL TAGS DESCRIPTION
langgraph/chatbot-agent Basic checkpointing Conversational AI c…
langgraph/mcp-agent Intermediate mcp MCP tool integratio…
langgraph/qa-agent Intermediate tool-calling Q&A assistant with …
langgraph/task-agent Intermediate hitl Human-in-the-loop t…
langgraph/tracing-agent Intermediate tracing OTEL distributed tr…
langgraph/research-agent Advanced http-policy Web research with f…
langgraph/coordinator-agent Expert multi-agent Multi-agent orchest…
langgraph/pentest-agent Expert shell Penetration testing…
crewai/research-agent Intermediate mcp,multi-agent Multi-agent CrewAI …
adk/research-agent Intermediate multi-agent,tool-calling Multi-agent researc…
strands/research-agent Intermediate mcp,checkpointing Research agent buil…
strands/graph-approval-agent Advanced multi-agent,checkpointing,hitl Two-node Strands Gr…
interactive/claude-code Expert shell,oauth Claude Code CLI run…
interactive/simple-shell Expert shell Interactive shell v…
interactive/vscode-tunnel Expert shell VS Code remote tunn…
clients/a2a-client Intermediate a2a Connect to AgentVis…
clients/mcp-client Intermediate mcp Connect to AgentVis…

Use: agentvisor template create <framework>/<name>

Filter by framework or tag:

agentvisor template list --framework crewai
agentvisor template list --tag mcp

show

Display the full README for a template before creating it.

agentvisor template show <framework>/<name>

Arguments

ArgumentDescription
<framework>/<name>Template identity, copied from the TEMPLATE column of template list

Examples

agentvisor template show langgraph/chatbot-agent
agentvisor template show crewai/research-agent

create

Scaffold a new agent project from a built-in template.

agentvisor template create <framework>/<name> [options]

Arguments

ArgumentDescription
<framework>/<name>Template identity, copied from the TEMPLATE column of template list

The framework prefix is always required, including for names that happen to be unique today — otherwise a working command would break the day another framework adds a template of the same name. create (and show) reject a bare name with the exact commands to run instead:

$ agentvisor template create research-agent
Error: template must be given as <framework>/<name>, got "research-agent"
Did you mean:
agentvisor template create adk/research-agent
agentvisor template create crewai/research-agent
agentvisor template create langgraph/research-agent

Options

OptionDefaultDescription
--forcefalseOverwrite existing directory
-o, --output-Output directory (default: ./<template-name>)

What Gets Created

Each template includes (LangGraph shown; other frameworks swap the framework-specific files, e.g. crewai.yaml/crew.py for CrewAI, agent.yaml for ADK):

<project>/
├── agent.py # LangGraph agent code
├── langgraph.json # Graph configuration
├── requirements.txt # Python dependencies
├── compose.yml # Development services (Temporal + Ollama)
├── agentvisor.yaml # AgentVisor runtime config (proxy credentials, etc.)
├── mav-agent-config.yaml # Agent-level config (a2a metadata, framework overrides)
├── policies/
│ ├── domain.yml # Access control policies
│ └── test.yml # Policy test cases
├── .dockerignore # Files excluded from container builds
└── README.md # Template documentation

The compose.yml is shared across all templates and starts Temporal (workflow orchestration) and Ollama (local LLM inference).

Examples

# Create in a directory named after the template
agentvisor template create langgraph/chatbot-agent
cd chatbot-agent

# The framework prefix selects between same-named templates
agentvisor template create crewai/research-agent

# Create in a custom directory
agentvisor template create langgraph/chatbot-agent --output ./my-chatbot

# Overwrite an existing project
agentvisor template create langgraph/chatbot-agent --force

Typical Workflow

# Scaffold, start services, and run
agentvisor template create langgraph/chatbot-agent
cd chatbot-agent
docker compose up -d
agentvisor serve .

See Also

  • Tutorial — Step-by-step walkthrough using template create
  • Examples — What each template demonstrates