Provider Configuration

Fresh

SOP-PROVIDER-001: Configure AI Providers

Field	Value
ID	SOP-PROVIDER-001
Version	1.0
Status	Active

Provider Comparison Matrix

Feature	Claude Code	Codex	OpenCode
Image attachments	New + resumed	New only	New + resumed
Session resume	--resume flag	exec resume <id>	--session flag
Read-only mode	--permission-mode plan	--sandbox read-only	--agent plan
Slash commands	Batch-mode only	Interactive TUI only	Not supported
Cost tracking	Full breakdown	Token counts only	Per-step costs
Model selection	Via Anthropic account	-m, --model flag	--model provider/model
Context operations	Merge, export, transfer	Merge, export, transfer	Merge, export, transfer
Thinking display	Streaming assistant messages	Reasoning tokens (o3/o4-mini)	Streaming text chunks

Claude Code (Primary)

Installation

npm install -g @anthropic-ai/claude-code
claude login

How Maestro Runs It

claude --print --verbose --output-format stream-json

Features

• Full cost breakdown per session
• Slash commands work in batch mode
• Session resume with --resume
• Image attachments on new and resumed sessions
• Read-only mode: --permission-mode plan

OpenAI Codex

Installation

npm install -g @openai/codex

Environment

OPENAI_API_KEY=sk-...

How Maestro Runs It

codex --json

YOLO mode is enabled by default in Maestro.

Features

• Reasoning tokens visible for o3/o4-mini models
• Model selection via -m, --model flag
• Image attachments on new sessions only

OpenCode (75+ LLM Providers)

OpenCode is the gateway to local and alternative LLMs.

Installation

npm install -g opencode

Ollama Configuration

{
  "$schema": "https://opencode.ai/config.json",
  "model": "ollama/qwen3:8b-16k",
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://localhost:11434/v1"
      },
      "models": {
        "qwen3:8b-16k": {
          "name": "Qwen3 8B",
          "tools": true
        }
      }
    }
  }
}

LM Studio Configuration

{
  "model": "lm-studio/your-model",
  "provider": {
    "lm-studio": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://127.0.0.1:1234/v1"
      }
    }
  }
}

llama.cpp Configuration

{
  "model": "llamacpp/your-model",
  "provider": {
    "llamacpp": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "http://127.0.0.1:8080/v1"
      }
    }
  }
}

Model ID Format

Always use provider_id/model_id format (e.g., ollama/qwen3:8b-16k).

Factory Droid

How Maestro Runs It

factory-droid -o stream-json

Multiple Claude Accounts

Use two or more Claude Code Max subscriptions by pointing agents at separate config directories:

# 1. Authenticate each account
# 2. Copy config
cp -a ~/.claude ~/.claude-personal

# 3. Symlink shared resources back to canonical
ln -s ~/.claude/projects ~/.claude-personal/projects
ln -s ~/.claude/settings.json ~/.claude-personal/settings.json

In Maestro agent configuration, set:

CLAUDE_CONFIG_DIR=/home/you/.claude-personal

DANGER

Never run both account agents on the same project simultaneously.

Verification Checklist

• [ ] Provider installed and authenticated
• [ ] Agent created in Maestro with correct directory
• [ ] Test message sent and response received
• [ ] Cost tracking visible (if supported)
• [ ] Session resume working

See Also

• Provider Comparison — Side-by-side feature table
• Settings — Global environment variables for API keys
• Multiple Accounts — Detailed multi-account setup