Paid Media Agent

Quick start · Deployment · Documentation · Contributing

_{Setup console · Report illustration with example figures. Presentation varies by runtime.}

Paid Media Agent helps you understand what changed across your ad accounts and decide what to do next. Ask questions in Slack or the terminal, generate performance reports, and prepare campaign changes for review.

It comes with ad platform integrations, analysis and reporting skills, and a paid-media wiki. Connect your accounts, add your company context, and deploy with Managed Deep Agents or on your own infrastructure. You choose the model.

What it does

• Compare campaign performance. Track spend, conversions, cost per lead, and budget pacing across connected accounts.
• Investigate changes. Find the campaigns behind a shift, check the supporting data, and flag gaps that could change the conclusion.
• Produce reports. Generate weekly or monthly summaries with charts, campaign tables, and recommendations. Download HTML, or PDF when the host has the rendering libraries installed.
• Prepare changes for review. Propose campaign updates with a reason and a plan for checking the result.

The model decides what to investigate. Code calculates the metrics and checks report figures against the source data. Large responses stay in files; the model receives summaries with the source, date window, and data-quality flags.

Live ad account changes are off by default. Enabling them requires a configured write policy, release checks, and approval from an authorized reviewer. Approval applies to the exact proposal reviewed; editing it requires a new approval.

Quick start

Requires Python 3.11+ and uv.

git clone https://github.com/langchain-ai/paid-media-agent.git
cd paid-media-agent
uv sync
uv run paid-media-agent setup

The setup command opens a local console for choosing a model, connecting ad accounts, and deploying. No frontend build is needed. The CLI exposes the same connection actions with JSON output if you prefer the terminal.

You can also ask your coding agent to guide setup:

Read AGENTS.md and .agents/skills/paid-media-onboarding/SKILL.md. Help me connect a model, connect my ad accounts, add my business context, and choose a deployment path.

Try it without model keys or ad accounts:

uv run paid-media-agent demo --with-proposal

The offline demo runs a scripted analysis and a simulated budget change against synthetic accounts. It doesn't call a model or touch live campaigns.

Once configured, ask a question or generate a report:

uv run paid-media-agent ask "Which campaigns had the largest increase in cost per lead last week?"
uv run paid-media-agent report --cadence weekly

Questions use your configured model. The report command runs without one, using supported campaign-performance adapters.

Accounts and business context

Connect the platforms you use: