Skip to content
Get packs

Guide · Agent skill

How to use the AutoGPT Agent Platform agent skill

Updated · written with AI from the project's own documentation and checked against our catalogue data

The autogpt-agents skill teaches your AI assistant how the AutoGPT platform works, so it can help you design, build and deploy autonomous agents. It is written for people who want to build agents as connected visual workflows, or who want to self-host the platform and extend it. After following this guide you will know how to install the skill, ask for it in plain language, and what you need before running anything.

AutoGPT Agent PlatformOrchestra-Research/AI-Research-SKILLs · MIT

Quick start

  1. 1Download and unzip the skill. Download the ZIP from this page and unzip it into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills.
  2. 2Check the requirements. Confirm you have Docker, Docker Compose, Node.js with npm, Git, the autogpt-platform repository, and your provider API keys.
  3. 3Clone and configure the platform. Clone the AutoGPT repository, enter autogpt_platform, and copy .env.example to .env.
  4. 4Start the backend and frontend. Run docker compose up -d --build, then start the frontend with npm install and npm run dev.
  5. 5Ask for an agent in plain language. Describe the agent, its trigger and its goal, such as a webhook-driven workflow or a scheduled run.
  6. 6Add credentials. In the platform, go to Profile > Integrations, then enter API keys or authorise OAuth for each service your agent uses.

What is an agent skill and what does this one do?

An agent skill is a folder of instructions and reference files that an AI assistant loads when your request matches its description. This skill is named autogpt-agents and its description covers building visual workflow agents, deploying persistent autonomous agents, and building multi-step AI automation systems.

Once loaded, the assistant knows how AutoGPT is structured. Agents are graphs of nodes connected by links, and each node runs a block such as an LLM call, a webhook, an integration, or a decision step. Agents can run continuously when triggered by a schedule or a webhook.

  • Explains the visual agent builder: add blocks, connect nodes, configure inputs, run the agent.
  • Covers trigger types: manual execution, webhook triggers, and scheduled execution.
  • Covers the developer side: the Forge toolkit, custom abilities, and benchmark testing.
  • Gives Docker-based setup steps for running the platform locally.

How do you install the autogpt-agents skill?

Download the ZIP from this page and unzip it. You can install it in one of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

  • Personal install: unzip into ~/.claude/skills/
  • Project install: unzip into the project's .claude/skills/ folder
  • Claude apps: Settings → Capabilities → Skills, then upload the skill

What do you need before you use it?

The skill describes a full platform install, so check these requirements first. The platform is heavy, so it is not suited to quick one-off tasks.

You will need Docker and Docker Compose, Node.js and npm for the frontend, Git, and the AutoGPT platform repository (autogpt-platform version 0.4.0 or later). You also need provider API keys such as OpenAI or Anthropic, and OAuth authorisation for integrations like GitHub or Google.

  • Docker and Docker Compose
  • Node.js and npm (frontend)
  • Git
  • AutoGPT platform repository, autogpt-platform>=0.4.0
  • Provider API keys such as OpenAI or Anthropic
  • OAuth for integrations such as GitHub or Google

How do you trigger the skill with natural requests?

You do not need a special command. Describe what you want in plain language and the assistant will load the skill when your request fits its description. Mention the kind of agent you are building, whether it should run continuously, and what should start it.

Good requests name the goal and the trigger. For example, you might ask how to set up a visual workflow agent that runs when a webhook fires, or how to deploy a persistent agent on a schedule.

  • "Help me design a visual workflow agent that summarises new GitHub issues."
  • "Walk me through running the AutoGPT platform locally with Docker."
  • "How do I trigger an AutoGPT agent with a webhook?"
  • "Create a scheduled AutoGPT agent that runs every two hours."

How do you run the platform locally?

The skill includes these setup steps for running the platform with Docker. Run them from a terminal. The backend starts with Docker Compose, and the frontend runs separately with npm.

Once both are running, the source lists the frontend UI at localhost port 3000, the backend API at localhost port 8006, and the WebSocket at port 8001.

  • Frontend UI: http://localhost:3000
  • Backend API: http://localhost:8006/api
  • WebSocket: ws://localhost:8001/ws
git clone https://github.com/Significant-Gravitas/AutoGPT.git cd AutoGPT/autogpt_platform cp .env.example .env docker compose up -d --build # In a separate terminal cd frontend cp .env.example .env npm install npm run dev

What can agents do and how are they triggered?

An agent is a graph. Its nodes run blocks, which are reusable components. The source lists block types including INPUT, OUTPUT, AI, WEBHOOK, STANDARD and AGENT. Agents can be started manually through the API, by a webhook, or on a schedule.

Scheduled runs use a cron-style expression, and monitoring can be done through WebSocket updates or by polling the execution endpoint.

  • Manual execution: POST to /api/v1/graphs/{graph_id}/execute with an inputs object.
  • Webhook: POST to /api/v1/webhooks/{webhook_id} with a payload.
  • Schedule: a cron expression such as 0 */2 * * * with a graph_id.
  • Monitoring: WebSocket updates, or GET /api/v1/executions/{execution_id}.

What about the developer toolkit and benchmarks?

The skill also covers AutoGPT Classic, which contains Forge, a toolkit for creating custom agents, and a benchmark system for testing agent performance. The source shows commands for setting up Forge, creating an agent from a template, and starting its server.

Benchmarks cover coding, retrieval, web and writing categories, and can record or replay HTTP responses (VCR cassettes) so results are reproducible.

  • Forge creates agents from a template and includes abilities, prompts and a config.yaml.
  • Benchmarks can be filtered by category or agent.
  • Use --record to create cassettes and --playback to reuse them.
cd classic ./run setup ./run forge create my-agent ./run forge start my-agent

What people use it for

  • Building a visual workflow that chains an LLM call with a decision step and an output node.
  • Running a persistent agent that starts when an external webhook sends data.
  • Connecting an agent to GitHub or Discord through integration blocks.
  • Developing a custom ability in Forge and testing it with the benchmark runner.

Frequently asked questions

Is the AutoGPT Agent Platform skill free?

The skill is published under the MIT licence. The source does not state prices for the AutoGPT platform or for the provider APIs it calls, so check those services directly before you run agents.

What do I need to install the autogpt-agents skill?

You need the skill ZIP from this page, unzipped into ~/.claude/skills/ or a project's .claude/skills/ folder, or uploaded under Settings → Capabilities → Skills in the Claude apps. Running the platform itself needs Docker, Docker Compose, Node.js, npm, Git and provider API keys.

Can a non-coder use AutoGPT?

The skill describes a visual drag-and-drop builder, so non-coders can build workflows without writing code. You still need to run the platform with Docker, which involves terminal commands, so some technical comfort helps.

How does AutoGPT compare to LangChain or CrewAI?

The skill itself says AutoGPT suits visual, persistent, trigger-driven agents. It points to LangChain or LlamaIndex when you need more control over agent logic, and to CrewAI for role-based multi-agent collaboration.

Why won't the platform start?

The troubleshooting reference covers common errors. For example, if Docker reports it cannot connect to the daemon, start Docker and confirm it with docker ps. If a port is already in use, find the process with lsof and stop it, or change the port in docker-compose.yml.

Are scheduled or webhook agents safe to leave running?

Continuous agents can take actions without review, and the skill does not describe safeguards. Use narrow credentials, test with manual runs first, and avoid giving broad access to unattended agents.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Choose this skill if you want to self-host the AutoGPT platform, build visual workflow agents, or connect persistent agents to services like GitHub or Discord. Skip it for simple one-off questions, small tasks, or if you only need a lightweight code library, where LangChain or CrewAI are the better fit.

Download the skill