Hi Product Hunt — I'm the maker of Sigoo.

If you remember the earlier SIGOO launch, that product converted videos into GIFs. Sigoo has since changed direction. This launch introduces Sigoo MCP Router, a tool-access and execution control layer for AI agents.

The problem it addresses is straightforward: connecting an agent to a tool should not automatically authorize every action that tool can perform.

Sigoo separates the workflow into four steps:

• Connect MCP capabilities behind one endpoint.

• Grant each agent access to specific tool versions.

• Check policies and require approval where configured before execution.

• Inspect an Action Receipt showing the outcome and its supporting evidence.

Agents use three Meta Tools to search permitted capabilities, execute a tool, and retrieve an action record. An uncertain result stays unknown rather than being presented as a success.

You can explore the no-login demo here:

https://sigoo.ai/demo

The demo uses fictional support data and illustrative receipts. It runs in the browser and does not connect accounts, create credentials, or call a real MCP server.

For protocol examples:

https://sigoo.ai/quickstart

Sigoo is in Founder Beta. The demo is not a production benchmark, and an Action Receipt is not a security certification or a guarantee of an external outcome.

If you're building agents that use external tools, where is the hardest part today: deciding what they can access, getting approvals right, or understanding what happened after execution?

I'd appreciate concrete workflows and technical feedback.