How to Set Up Your AI Marketing Agent with Codex [2026]

In Part 1, we covered why an AI marketing agent is different from an AI assistant — persistent brand knowledge, strategy memory, specialized skills, autonomous scheduling. Now we build it.

The setup takes about 10 minutes. Most of that time is the brand workspace creation, where the agent scans your brand docs and pulls live data from your ad accounts to build its knowledge base. The actual installation is one command.

Prerequisites

Before you start, make sure you have:

• OpenAI Codex CLI installed (installation guide)
• Git installed (the installer clones from GitHub)
• Node.js 18+ (for the MCP server)
• An Adspirer account (free to start)
• At least one ad platform account: Google Ads, Meta Ads, LinkedIn Ads, or TikTok Ads
• A ChatGPT Plus, Pro, Business, or Enterprise subscription (Codex requires an OpenAI paid plan)

Step 1: Install the Adspirer Plugin

The one-command installer sets up everything — the MCP server connection, all 5 skills, the agent config, and safety rules:

bash <(curl -fsSL https://raw.githubusercontent.com/amekala/ads-mcp/main/plugins/codex/adspirer/install.sh)

This installs:

After installation, restart Codex to load the new configuration.

# Add MCP server
codex mcp add adspirer --url https://mcp.adspirer.com/mcp

# Clone and install skills + agent
git clone https://github.com/amekala/ads-mcp.git /tmp/ads-mcp
mkdir -p ~/.agents/skills
cp -r /tmp/ads-mcp/plugins/codex/adspirer/skills/adspirer-ads ~/.agents/skills/
cp -r /tmp/ads-mcp/plugins/codex/adspirer/skills/adspirer-setup ~/.agents/skills/
cp -r /tmp/ads-mcp/plugins/codex/adspirer/skills/adspirer-performance-review ~/.agents/skills/
cp -r /tmp/ads-mcp/plugins/codex/adspirer/skills/adspirer-write-ad-copy ~/.agents/skills/
cp -r /tmp/ads-mcp/plugins/codex/adspirer/skills/adspirer-wasted-spend ~/.agents/skills/

# Install agent config
mkdir -p ~/.codex/agents
cp /tmp/ads-mcp/plugins/codex/adspirer/agents/performance-marketing-agent.toml ~/.codex/agents/

[features]
multi_agent = true

[agents.performance-marketing-agent]
description = "Brand-specific performance marketing agent."
config_file = "agents/performance-marketing-agent.toml"

Step 2: Authenticate with Adspirer

Verify the MCP server is installed:

codex mcp list

You should see adspirer in the output (it may show “Unsupported” for auth — that’s normal before login). Now authenticate:

codex mcp login adspirer

A browser window opens. Sign in with your Adspirer account (or create one free) and link your ad platform accounts — Google Ads, Meta Ads, LinkedIn Ads, and/or TikTok Ads. The connection uses OAuth 2.1, and your ad account credentials are never stored locally. See the security documentation for details.

After authenticating, verify:

codex mcp list

adspirer should now show as enabled.

Step 3: Create Your Brand Workspace

This is where the agent becomes yours. Navigate to your brand’s project folder and launch Codex:

cd ~/Clients/YourBrand    # or wherever your brand docs live
codex

Then tell the agent to set up your workspace:

Or invoke the skill directly with $adspirer-setup.

What Happens During Setup

The setup skill runs a comprehensive bootstrap process:

What Your Workspace Looks Like After Setup

~/Clients/YourBrand/
├── AGENTS.md           # Brand context — everything the agent knows about you
├── STRATEGY.md         # Strategy directives — persists across sessions
├── brand-guidelines.pdf  # Your existing docs (if you added any)
└── media-plan.xlsx       # Your existing docs

The AGENTS.md file is the agent’s brain. Here’s what a typical one contains:

Every time the agent runs a task, it reads AGENTS.md first. This is how it knows your brand without you re-explaining it every session.

Step 4: Understand the 5 Skills

Your agent has 5 specialized skills — each is a proven workflow that enforces the right tool sequence for a specific task type. You don’t need to remember their names; describe what you want and Codex matches the right skill automatically.

For the complete skill reference with all parameters and workflows, see the skill documentation.

Step 5: Safety Rules

When an AI agent operates autonomously, safety is critical. Codex’s .rules files enforce safety at a deeper level than skill instructions — they block dangerous operations before they reach the AI.

What’s Protected

Step 6: Your First Autonomous Workflow

Let’s set up the most valuable autonomous workflow — a daily performance check that runs every morning without you opening any tool.

Configure the Daily Check

Edit your agent’s default prompt in ~/.codex/agents/performance-marketing-agent.toml:

[agent]
name = "performance-marketing-agent"
description = "Daily ad performance monitoring and analysis"

[agent.default_prompt]
text = """
Pull yesterday's performance across all my connected ad platforms.
Compare each campaign to the previous 7-day average.
Flag any campaigns where:
- CPA increased more than 30%
- Spend increased more than 50% without proportional conversion increase
- CTR dropped below 1%
- Frequency exceeded 3.0 (Meta/LinkedIn)

Check STRATEGY.md for any active directives that might affect recommendations.
Give me a prioritized summary with the top 3 actions I should take today.
"""

Run It

codex run performance-marketing-agent

The agent reads AGENTS.md for brand context, pulls performance data from all connected platforms via Adspirer, compares against your KPI targets, checks STRATEGY.md for strategic directives, and produces a prioritized summary.

What You Get

The agent returns a structured performance report — no dashboard clicking, no CSV exports, no spreadsheet building. Every metric compared to your targets, every anomaly explained, every recommendation backed by data.

And tomorrow morning, it runs again automatically.

Approval Modes

Codex has 3 approval modes that control how much autonomy the agent gets:

For advertising, Auto mode is the sweet spot. Read operations (pulling performance data, analyzing campaigns) run automatically. Write operations (creating campaigns, changing budgets) prompt you for confirmation. This matches the safety model — the agent handles analysis autonomously but defers to you for spending decisions.

Troubleshooting

What’s Next

Your agent is installed, authenticated, and running. The brand workspace is created. You’ve got 5 skills, safety rules, and your first autonomous workflow configured.

In Part 3: How to Run Facebook & Instagram Ads with Your Codex AI Agent, we go deeper — building a Meta Ads specialist role with creative fatigue monitoring, audience optimization, campaign ideation pipelines, and the compounding knowledge effect that makes the agent smarter over time.

Get started free →

Related Articles

• How to Build an AI Marketing Agent for Paid Media — Part 1: The concept and why it matters
• How to Run Facebook & Instagram Ads with Your Codex AI Agent — Part 3: Channel-specific role development
• How to Run Facebook & Instagram Ads with ChatGPT — The assistant approach
• How to Run Facebook & Instagram Ads with Claude AI — Claude’s deep reasoning approach
• What Is MCP (Model Context Protocol)? — The protocol behind all of this
• PPC Automation with ChatGPT and Claude — Automating paid search workflows