Moltis uses TOML configuration files (not environment variables) for all settings. The primary configuration file is moltis.toml, located in ~/.config/moltis/ by default.
| Platform | Config Path | Data Path |
|---|---|---|
| Linux | ~/.config/moltis/moltis.toml |
~/.moltis/ |
| macOS | ~/.config/moltis/moltis.toml |
~/.moltis/ |
| Docker | /home/moltis/.config/moltis/moltis.toml |
/home/moltis/.moltis/ |
# moltis.toml - Core configuration
[server]
# Bind address (default: localhost for local, requires auth for remote)
bind = "127.0.0.1:13131"
# HTTPS configuration (recommended for production)
# https.cert = "/path/to/cert.pem"
# https.key = "/path/to/key.pem"
[paths]
# Data directory for sessions, memory, and agent state
data = "~/.moltis"
# Configuration directory
config = "~/.config/moltis"
[sandbox]
# Sandbox runtime: "docker", "podman", or "apple" (macOS only)
runtime = "docker"
# Optional: Default resource limits for sandboxed containers
# memory_limit = "512MB"
# cpu_limit = 1.0
[memory]
# Hybrid memory configuration
# Enables vector + full-text search for long-term context
enabled = true
# Embedding provider for vector search
# embedding_provider = "openai" # or "local" for local GGUF embeddings
[providers]
# LLM Provider configuration
# Add your API keys or local model paths
[providers.openai]
api_key = "sk-..." # Or use MOLTIS_PASSWORD env var for secrets
[providers.anthropic]
api_key = "sk-ant-..."
[providers.ollama]
# Local Ollama instance
endpoint = "http://localhost:11434"
[mcp]
# Model Context Protocol servers
# Enable MCP server connections (stdio or HTTP/SSE)
[[mcp.servers]]
name = "example-server"
command = "npx"
args = ["-y", "@example/mcp-server"]
# Or for HTTP/SSE:
# url = "http://localhost:8080/sse"
[channels]
# Enable communication channels
[channels.telegram]
# Telegram bot token
# token = "BOT_TOKEN"
[channels.discord]
# Discord bot token
# token = "BOT_TOKEN"
[voice]
# Voice I/O configuration (15+ TTS/STT providers)
[voice.tts]
# Text-to-speech provider
# provider = "openai"
[voice.stt]
# Speech-to-text provider
# provider = "openai"
[security]
# Security settings
# Require authentication for non-localhost access
require_auth_remote = true
# Rate limiting (requests per minute per IP)
rate_limit = 60
# SSRF protection (blocks internal network requests)
ssrf_protection = true
Moltis primarily uses TOML configuration. Only these environment variables are recognized:
# Password for cloud/remote deployments (overrides config file)
MOLTIS_PASSWORD=your_secure_password
127.0.0.1:13131) unless remote access is requireddocker, podman, or apple)MOLTIS_PASSWORD env var for secrets| Port | Service |
|---|---|
| 13131 | HTTPS Web UI (primary) |
| 13132 | Additional service |
| 1455 | Additional service |
Back up the following directories:
~/.config/moltis/ - Configuration files (moltis.toml)~/.moltis/ - Data directory (sessions, memory, agent state, SQLite databases)Validate restores by:
Any questions?
Feel free to contact us. Find all contact information on our contact page.