forge

AutoGPT Forge

Core autonomous agent framework for building AI agents.

Quick Start

All commands run from the classic/ directory (parent of this directory):

# Install (one-time setup)
cd classic
poetry install

# Configure
cp .env.example .env
# Edit .env with your OPENAI_API_KEY

# Run
poetry run python -m forge

The agent server runs on http://localhost:8000 by default.

Configuration

Environment Variables (.env)

# Required
OPENAI_API_KEY=sk-...

# Optional LLM settings
SMART_LLM=gpt-4o                    # Model for complex reasoning
FAST_LLM=gpt-4o-mini                # Model for simple tasks
EMBEDDING_MODEL=text-embedding-3-small

# Optional search providers
TAVILY_API_KEY=tvly-...
SERPER_API_KEY=...
GOOGLE_API_KEY=...
GOOGLE_CUSTOM_SEARCH_ENGINE_ID=...

# Optional infrastructure
LOG_LEVEL=DEBUG                     # DEBUG, INFO, WARNING, ERROR
DATABASE_STRING=sqlite:///agent.db  # Agent Protocol database
PORT=8000                           # Server port
FILE_STORAGE_BACKEND=local          # local, s3, or gcs

Workspace Settings (.autogpt/autogpt.yaml)

Workspace-wide permissions for all agents:

allow:
  - read_file({workspace}/**)
  - write_to_file({workspace}/**)
  - list_folder({workspace}/**)
  - web_search(*)

deny:
  - read_file(**.env)
  - read_file(**.key)
  - execute_shell(rm -rf:*)
  - execute_shell(sudo:*)

Agent Settings (.autogpt/agents/{id}/permissions.yaml)

Agent-specific permission overrides:

allow:
  - execute_python(*)
deny:
  - execute_shell(*)

Workspace Structure

{workspace}/
├── .autogpt/
│   ├── autogpt.yaml              # Workspace permissions
│   ├── ap_server.db              # Agent Protocol database
│   └── agents/
│       └── AutoGPT-{agent_id}/
│           ├── state.json        # Agent state
│           ├── permissions.yaml  # Agent permissions
│           └── workspace/        # Agent's working directory

Permissions

Permission checks follow this order (first match wins):

1. Agent deny list → Block
2. Workspace deny list → Block
3. Agent allow list → Allow
4. Workspace allow list → Allow
5. Prompt user → Interactive approval

Pattern Syntax

Format: command_name(glob_pattern)

Pattern	Description
read_file({workspace}/**)	Read any file in workspace
execute_shell(python:**)	Execute Python commands
web_search(*)	All web searches

Special tokens:

• {workspace} - Replaced with workspace path
• ** - Matches any path including /
• * - Matches any characters except /

Tutorials

The tutorial series guides you through building a custom agent:

1. A Comprehensive Guide to Your First Steps
2. The Blueprint of an AI Agent
3. Interacting with your Agent
4. Crafting Intelligent Agent Logic