Notion Adds Usage Controls and Analytics for Custom Agents.

Notion rolled out new administrative controls and visibility features for Custom Agents, addressing cost management concerns as teams scale their use of autonomous AI workflows. The update, announced on LinkedIn, introduces permission controls, per-agent credit limits, and a centralized credits dashboard.

The changes come months after Notion launched Custom Agents to automate complex workflows like triaging support tickets, enriching CRM data, and summarizing meeting notes. As adoption grew, workspace admins needed better tools to monitor usage and prevent runaway costs.

Admins can now restrict who has permission to create agents within their workspace. The control lives in Settings under Notion AI and Agents. Once toggled, only authorized users can build new agents, giving teams a way to gate AI deployment.

Each agent can also be assigned a credit limit at creation. When an agent hits its cap, it pauses automatically. Notion sends notifications as agents approach their limits and when they pause due to exhausted credits.

The update also adds an automatic safety rail that detects unusual spend spikes. If an agent suddenly consumes credits at an unexpected rate, Notion will pause it and alert the admin or creator. This catch prevents scenarios where a misconfigured agent drains a workspace's credits overnight.

The new credits dashboard breaks down usage by individual agent. Admins can see which agents are consuming the most credits, identify trends over time, and disable agents that are pacing higher than expected.

The interface is designed to give workspace owners a clear view of what's costing what, without requiring them to dig through logs or wait for monthly billing summaries.

Later this month, Notion plans to ship deeper analytics for both admins and agent creators. These will include per-run cost, success rate, and how often teammates are invoking each agent. The goal is to help teams troubleshoot underperforming agents and optimize workflows based on actual usage patterns.

Custom Agents represent a shift in how teams use AI tools. Unlike traditional SaaS features with predictable per-seat pricing, agents operate autonomously and consume variable compute resources. A poorly scoped agent or an unexpected volume spike can lead to surprise costs.

By adding granular controls before spend happens, Notion is addressing a core concern for enterprise buyers evaluating agentic AI. The analytics layer also gives teams data to justify continued investment or identify which use cases deliver the most value.

Notion has been expanding its AI capabilities steadily, including integrations with external models and tools like Grok via MCP. The platform is positioning itself as an operating layer where teams can deploy AI across workflows, not just as a writing assistant.

The new controls are available now to all workspaces using Custom Agents. The deeper analytics features are expected to roll out later in May.