Agent Planners: The Approval-Gated Multi-Agent Marketing Harness

Research context and background

Agent Planners is a smart system designed to help marketing teams plan and run strategies for search engine optimization, data analysis, advertising, and publisher platforms. Unlike other AI tools that make changes automatically without anyone watching, Agent Planners works in the open. It shows users exactly what it plans to do and waits for a human to say yes before making any changes to live accounts. This tool connects directly to Google products like Google Ads and Analytics as well as over 250 other apps. It offers a free plan that gives users 2,500 credits every month without needing a credit card to start.

Benefits

Agent Planners offers several key advantages for marketing teams who want to use AI safely. First, every action is visible. The system creates a step-by-step plan that explains its reasoning before it runs anything. Second, no changes happen without approval. Users can sign off on tasks inside the app, via a special email link, or by pre-approving specific types of work. If one step is rejected, the rest of the plan can still continue. Third, everything is fully recorded. The platform logs every change with details on who made it, when it happened, why it was done, and how the account looked before and after. This creates a clear history for compliance and accountability. The system also uses a team of six specialized agents that handle different jobs like analytics, creative work, and safety checks. Users can choose which AI model to use for these tasks and even bring their own models to save money. The tool also remembers past successes and learns from them to improve future performance.

Use Cases

Agent Planners can be used in many ways depending on the needs of the marketing team. For read-only analysis, it can summarize weekly performance data, flag campaigns with rising costs, and monitor ads that might violate policies. It can also review account change histories to see who made what adjustments. For optimization, the tool can suggest pausing campaigns that do not meet profit targets, find search terms that waste money, and recommend shifting budgets to better-performing ads. It can also help launch new campaigns by creating brand search ads or building performance-based campaigns with generated images and videos. For SEO and publisher revenue, it tracks ranking drops, finds keywords that need new content, and reviews earnings from ad networks. The platform can also automate cross-app tasks like syncing customer data to a CRM, sending real-time alerts to Slack when budgets spike, or automatically generating weekly reports for executives. It supports managing multiple accounts at once and can run tasks on a schedule to keep things updated.

Pricing

Agent Planners offers a free tier that includes 2,500 credits per month. No credit card is required to start using the platform. For users who need more power, the system supports a Bring Your Own Model feature. This allows users to connect their own AI models through providers like Azure AI or OpenRouter. Tokens used on these external models do not cost platform credits, so users only pay their chosen provider directly while using Agent Planners for the orchestration and tooling.

Vibes

While specific customer reviews are not detailed in the available information, the platform is built on a philosophy that values safety and transparency. The core idea is that autonomy without visibility is a risk. By requiring human approval for every change and providing full audit trails, the tool aims to build trust between AI automation and human oversight. The design focuses on reliability, ensuring that the system checks its work before acting and can fix errors on its own without stopping the entire process.

Additional Information

Agent Planners is built on a multi-agent architecture where a central controller manages a team of specialists. Each specialist handles a specific function like analytics, optimization, creative work, research, safety, and reporting. The system supports dynamic switching between different AI models without changing the workflow. It also features a skills library that saves successful methods so the agents can repeat proven approaches instead of starting from scratch every time. The platform includes security measures like encrypted credentials and scoped permissions to control access within an organization. Users can define roles such as Admin, Editor, or Member to manage who has what level of control. Integrations and access keys can be revoked at any time to maintain security.