“I don’t want to talk to your agents in your tool. I want my agent to talk to your tool.” That quote perfectly sums up how our way of working has changed over the past few months. Today, we work almost entirely from Claude: - Keyword research using the DataForSEO and SEO Gets MCPs. - Technical SEO by controlling Screaming Frog via MCP, or running it through CLI on our VPS with a Claude managed agent. - Data analysis using SEO Gets, Gscwizard.com, Rankshift, and our own tracking tools. - Content creation connected to AirOps or our own knowledge base, combined with tools like Firecrawl and Parallel Web Systems. - Time tracking through an unofficial Teamleader MCP. Everyone on our team has built their own workflows, connected their preferred MCPs, and set things up in the tools they like working with. Yet I still see so many companies building their own AI agents, adding workflows inside their platforms, and trying to become the next all-in-one solution. We're moving in the opposite direction. That's why we built Depends OS. Not another AI agent, but one central place where everything comes together: - All client brand information, context, and relevant sources. - Playbooks and skills that can be customized per client and content type. - Connected tools and data sources, so clients don't need their own DataForSEO account or have to configure a GSC MCP. - Content that anyone on the team can review, edit, and manage, including through MCP. But here's the important part: Almost nobody has to work inside Depends OS itself. Everyone can connect to Depends OS via MCP, directly from their own Claude environment. They can brainstorm with it, retrieve client context, combine data from other MCPs, create content, or edit existing articles. All without leaving the environment they prefer. And I think that's exactly where the challenge lies for many software companies. You can never build features and workflows fast enough to keep up with how creative your users are. But you can give them the right context, tools, and capabilities to build their own ideal workflows. Instead of building another tool with an AI agent inside, build a tool that any AI agent can work with. Curious how others are approaching this. Are you still working mainly inside individual tools, or are you increasingly using your own AI agent as the central interface?
