OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Frameworks

Agent Starter Kit

Agent Starter Kit is an MIT-licensed Shell framework that uses Markdown files to coordinate AI coding personas and multi-model workflows from an AI command-line interface. Its Maestro can decompose complex requests and dispatch work to specialized roles, while quick mode accepts direct plain-language tasks.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Agent Starter Kit — OpenViST project profile in Agent Frameworks.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when an AI CLI needs project-local instructions and a repeatable multi-agent workflow for software tasks. Quick mode is suited to direct requests, while full orchestration is suited to multi-step work that benefits from decomposition, dispatch and review.

Who it is for

  • Developers organizing AI-assisted work inside software repositories
  • Teams extending a reusable set of personas, rules and skills
  • Users of AI command-line interfaces who want model-agnostic orchestration

Documented uses

  • Turn plain-language software requests into quick or staged AI-assisted workflows
  • Divide complex work among planning, coding, review and context-mapping personas
  • Keep project-specific agent instructions, procedures and rules in a versioned scaffold
  • Route persona work across selected language models or use one model for all roles
  • Retain corrections, preferences and session information for later work
Before you choose

Requirements & limits

The repository is public and identifies an MIT license. Setup is performed inside each software project and requires an AI agent interface; OpenCode-specific automatic binding also depends on yq and jq being installed.

  • The framework is a foundation and starter scaffold, so domain-specific behavior is expected to be added in a fork or project customization.
  • Its core configuration is Markdown-based and uses project-local files rather than a separately installed plugin.
  • Full orchestration uses distinct Architect, Coder, Reviewer and Contextualizer roles, while quick mode bypasses the Maestro flow.
  • The repository describes compatibility with AI CLIs that accept prompts through stdin; OpenCode has native persona binding described in the setup documentation.
  • Model routing is presented as model-agnostic, but the evidence does not document a formal provider API or exact configuration mechanism.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Agent Starter Kit in a software project

Install the project-local scaffold and make its instructions available to an AI CLI.

Before you start

  • A target software project
  • Git access to https://github.com/ntorga/agent-starter-kit
  • An AI agent interface such as Claude, OpenCode or another CLI/TUI that accepts prompts via stdin

Steps

  1. From the target project, clone the repository into a .agents directory.

  2. Create a root-level symlink from .agents/AGENTS.md to AGENTS.md.

  3. Start the AI agent interface.

  4. For full orchestration, begin with a request to comply with @.agents/ENTRYPOINT.md; for quick work, describe the task directly.

Expected outcome

The project contains the framework scaffold and exposes its AGENTS.md instructions at the project root for the AI CLI.

Guide limitations

  • The setup instructions describe project-local copies rather than a single system-wide installation.
  • The evidence does not document the result of running a task after setup.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
3 October 2026
Original website title
ntorga/agent-starter-kit · Official homepage title · retained in its source language
How we review sourcesSuggest a correction