Make Launches MCP Toolboxes to Give AI Agents Controlled Access to Workflows.

Make announced MCP Toolboxes, a new feature that lets teams safely connect AI clients like Claude, Cursor, and ChatGPT to their automation workflows with granular access control. The feature addresses a growing challenge as AI agents increasingly need to execute actions across business systems, not just provide information.

The toolboxes solve a fundamental tension in AI orchestration: giving agents enough access to be useful without exposing entire API surfaces or creating governance blind spots. Rather than connecting an AI directly to platforms where it might accidentally delete critical data or waste tokens parsing hundreds of available functions, Make lets teams select specific scenarios to expose as callable tools.

Teams create toolboxes at the organization level in Make, then populate them with selected scenarios from their existing workflows. Each toolbox generates a unique URL and authorization token that connects to compatible AI clients using the Model Context Protocol standard.

The setup process is straightforward. After creating a toolbox and assigning scenarios, teams copy the toolbox URL into their AI client's MCP configuration. From that point, the agent can only trigger the workflows explicitly included in that toolbox, nothing more.

Organizations can create multiple toolboxes with different scenario collections for various agents, teams, or use cases. A support team's agent might access customer data lookup and ticket creation workflows, while a sales agent connects only to CRM update and lead routing scenarios. Each operates independently with its own authorization token.

Make designed the feature around production environment needs: deterministic execution, observability, and managed governance. The company's blog post notes that native MCP connections often expose entire API surfaces and force language models to infer multi-step workflows, inflating token usage and increasing the risk of incorrect actions.

Toolboxes provide centralized visibility into which tools are being called, by which agents, and how often. Organizations can revoke access by rotating tokens or removing scenarios from a toolbox without changing the underlying workflows. The scenarios themselves remain the same; only their availability to specific AI clients changes.

Make positions the feature as part of its broader AI agent strategy. The company recently launched an AI Agent Library with pre-built templates and introduced public sharing for AI agent scenarios, building out infrastructure for teams to build and share agentic workflows.

One technical benefit MCP Toolboxes addresses is context window efficiency. When an AI client connects directly to a large platform, it receives specifications for potentially hundreds of available functions. The language model must parse all those options every time it considers an action, consuming tokens without adding value.

By limiting the tool set to a curated collection of scenarios, Make reduces the amount of information the model processes for each decision. The agent sees only relevant, pre-configured workflows rather than raw API endpoints, making execution faster and more predictable while lowering usage costs.

The feature is available now at the team level in Make. Organizations interested in connecting AI agents to their automation workflows can create toolboxes through their Make account settings and configure them with existing scenarios.