Configuration

Configure AgenticX for your environment.

Configuration

AgenticX uses ~/.agenticx/config.yaml as its global configuration file.

Global Config

yaml

1# ~/.agenticx/config.yaml
2
3# Default LLM provider
4default_provider: openai
5default_model: gpt-4o
6
7# Runtime settings
8AGX_MAX_TOOL_ROUNDS: 20
9
10# Studio server
11studio:
12  host: 0.0.0.0
13  port: 8000
14
15# Memory settings
16memory:
17  backend: sqlite   # sqlite | redis | postgresql
18  path: ~/.agenticx/workspace

Provider Configuration

Configure LLM providers in config or via environment variables:

OpenAI

yaml

1providers:
2  openai:
3    api_key: ${OPENAI_API_KEY}
4    default_model: gpt-4o

Anthropic

yaml

1providers:
2  anthropic:
3    api_key: ${ANTHROPIC_API_KEY}
4    default_model: claude-3-5-sonnet-20241022

Ollama (local)

yaml

1providers:
2  ollama:
3    base_url: http://localhost:11434
4    default_model: llama3.2

MiniMax

yaml

1providers:
2  minimax:
3    api_key: ${MINIMAX_API_KEY}
4    default_model: minimax-m1

Environment Variables

Variable	Description	Default
`OPENAI_API_KEY`	OpenAI API key	—
`ANTHROPIC_API_KEY`	Anthropic API key	—
`AGX_MAX_TOOL_ROUNDS`	Max tool call rounds per agent turn	`20`
`AGX_CHROMIUM_QUIET`	Suppress Chromium logs in Desktop	`false`

Workspace & Identity

Agent identities and memory are stored in ~/.agenticx/workspace/. Each avatar gets its own subdirectory with session history, memories, and context files.

Extensions & Skill Marketplace

Configure extension registries and local scan directories for the skill marketplace:

yaml

1extensions:
2  registries:
3    - name: official
4      url: https://registry.agxbuilder.com
5      type: agx                          # AgenticX native registry
6    - name: clawhub
7      url: https://clawhub.ai/api
8      type: clawhub                      # ClawHub skills market
9  scan_dirs:
10    - ~/.agenticx/bundles
11    - ~/.agenticx/skills/registry

See Extensions & Skill Ecosystem for full details on AGX Bundles and marketplace usage.

Project-level Config

For per-project settings, create agenticx.yaml in your project root:

yaml

1organization_id: my-org
2default_llm:
3  provider: openai
4  model: gpt-4o-mini
5tools:
6  mcp_servers:
7    - name: filesystem
8      command: npx
9      args: ["-y", "@modelcontextprotocol/server-filesystem", "."]