AutoGPT Plugins and Community Ecosystem

Q: Are plugins from the old AutoGPT still compatible?

The classic plugin system (ALLOWLISTEDPLUGINS) has been deprecated in favor of Blocks and the Forge SDK. Old plugins need to be ported. Check the plugin-migration Discord channel for community-maintained ports.

The AutoGPT Ecosystem

AutoGPT is one of the most starred AI projects on GitHub (170,000+ stars). Around it has grown an ecosystem of plugins, custom agents, frontends, and community tools that dramatically extend what the platform can do.

This guide covers how to use existing plugins, where to find community resources, and how to build your own extensions.

The AutoGPT Marketplace

The AutoGPT Store (accessible from the AutoGPT web interface) lets you:

Browse agents built by the community
Run published agents without any setup
Publish your own agents for others to use

To access:

Go to beta.agpt.co (AutoGPT’s hosted platform)
Click “Agent Store” in the navigation
Browse by category: coding, research, productivity, data analysis

Community agents in the store are built with AutoGPT Forge (see AutoGPT Forge guide) and implement the Agent Protocol, so they work with any compatible frontend.

AutoGPT Blocks (Visual Builder)

AutoGPT’s visual builder uses Blocks — composable units that you wire together, similar to n8n nodes:

Available block categories:

AI/LLM blocks: Call any LLM model
Web blocks: Search, scrape, browse
File blocks: Read/write files, parse documents
Integration blocks: GitHub, Slack, Twitter/X, email
Data blocks: Filter, transform, aggregate

Example: Research + Summarize Block Flow

Web Search Block → Extract Content Block → Summarize with LLM Block → Write to File Block

To use Blocks:

Open AutoGPT at localhost:3000
Click “Build” → “New Agent”
Drag blocks from the palette and connect them

Community Plugins (Classic)

The classic AutoGPT supported plugins via a plugin directory. While the architecture has evolved toward Blocks, many community plugins are being ported:

Popular Community Extensions

Search and Research

Bing Search — Microsoft Bing integration (better than DuckDuckGo for current events)
Wikipedia — Query Wikipedia for factual information
ArXiv — Search academic papers

Productivity

Google Sheets — Read/write spreadsheet data
Notion — Integrate with Notion databases
Trello/Linear — Project management actions

Code and Development

GitHub — Create issues, PRs, push code
Replit — Run code in cloud sandboxes
Docker — Execute commands in isolated containers

Communication

Slack — Post messages and read channels
Discord — Bot integration for Discord servers
Email (SMTP) — Send and receive emails

Extending AutoGPT with Blocks

To add a custom integration as a Block:

# autogpts/autogpt/autogpt/blocks/my_integration.py
from autogpt.core.agents.simple import Block, BlockInput, BlockOutput

class NotionQueryBlock(Block):
    """Query a Notion database and return results."""

    name = "notion_query"
    description = "Query a Notion database by filter criteria"
    input_schema = {
        "database_id": {"type": "string", "description": "Notion database ID"},
        "filter": {"type": "string", "description": "Filter query"},
    }
    output_schema = {
        "results": {"type": "array", "description": "Matching Notion pages"},
    }

    async def execute(self, input: BlockInput) -> BlockOutput:
        import httpx
        headers = {
            "Authorization": f"Bearer {self.settings.notion_token}",
            "Notion-Version": "2022-06-28",
        }
        async with httpx.AsyncClient() as client:
            response = await client.post(
                f"https://api.notion.com/v1/databases/{input.database_id}/query",
                headers=headers,
                json={"filter": {"property": "Name", "rich_text": {"contains": input.filter}}},
            )
            data = response.json()
        return {"results": data.get("results", [])}

Community Resources

Official Channels

Resource	URL	Purpose
GitHub	github.com/Significant-Gravitas/AutoGPT	Source code, issues
Discord	discord.gg/autogpt	Community chat
Docs	docs.agpt.co	Official documentation

Community Highlights

Benchmarks and Comparisons The AutoGPT team runs regular benchmarks comparing agent performance on standardized tasks. The AgentBench leaderboard includes AutoGPT alongside competitors.

AutoGPT Arena Community platform where agents compete on tasks. Submit your Forge-built agent to see how it ranks.

AutoGPT Awesome List The community maintains awesome-autogpt — a curated list of projects, tutorials, and integrations.

Using AutoGPT’s API

AutoGPT exposes a REST API you can call programmatically:

import httpx
import asyncio

async def run_autogpt_task(task: str) -> str:
    base_url = "http://localhost:8080/ap/v1"

    # Create task
    async with httpx.AsyncClient() as client:
        resp = await client.post(
            f"{base_url}/agent/tasks",
            json={"input": task},
        )
        task_id = resp.json()["task_id"]

        # Execute steps until done
        max_steps = 20
        for _ in range(max_steps):
            step_resp = await client.post(
                f"{base_url}/agent/tasks/{task_id}/steps",
                json={},
            )
            step = step_resp.json()
            print(f"Step output: {step.get('output', '')[:200]}")

            if step.get("is_last"):
                return step.get("output", "")

    return "Max steps reached"

# Run
result = asyncio.run(run_autogpt_task(
    "Research the top 3 AI frameworks released in 2025 and summarize their key features"
))
print(result)

Contributing to AutoGPT

The AutoGPT project welcomes contributions:

Good first contributions:

Documentation improvements
New Block implementations
Bug fixes (check issues labeled good first issue)
Adding tests for existing blocks

Contribution process:

# Fork and clone
git clone https://github.com/YOUR_USERNAME/AutoGPT.git

# Create a feature branch
git checkout -b feature/my-new-block

# Make changes, run tests
poetry run pytest tests/

# Submit PR
gh pr create --title "Add Notion query block" \
  --body "Implements a Block for querying Notion databases"

The team reviews PRs actively — community blocks with good test coverage are typically merged within 1–2 weeks.

Frequently Asked Questions

Are plugins from the old AutoGPT still compatible?

The classic plugin system (ALLOWLISTED_PLUGINS) has been deprecated in favor of Blocks and the Forge SDK. Old plugins need to be ported. Check the #plugin-migration Discord channel for community-maintained ports.

Can I use AutoGPT plugins with other agent frameworks?

Not directly — AutoGPT Blocks and Forge abilities are specific to AutoGPT’s architecture. However, since Forge agents implement the Agent Protocol, any Agent Protocol-compatible orchestrator can call them as sub-agents.

How do I report a bug in a community plugin?

Open a GitHub issue in the main AutoGPT repo with the label plugin-bug. Include the plugin name, n8n/Forge version, error message, and steps to reproduce.

Is there a plugin for connecting AutoGPT to n8n?

Yes — you can call any n8n webhook from an AutoGPT HTTP Block. Create an n8n workflow with a webhook trigger, then use AutoGPT’s HTTP Request Block to invoke it. This lets you combine AutoGPT’s reasoning with n8n’s 400+ integrations.

How do I publish my agent to the AutoGPT Store?

Build your agent with Forge, ensure it passes the benchmark suite, then submit via the Store submission form at beta.agpt.co/store/submit. The AutoGPT team reviews for quality and security before publishing.

Next Steps

AutoGPT Forge: Build Custom Agents — Build your own Forge agent to publish
AutoGPT Use Cases — See real-world applications built with AutoGPT
CrewAI Multi-Agent Workflows — Compare with CrewAI’s approach to extensible agents