We Built Our Support Platform to Run Entirely From Claude. Here's How — and Why It Matters.

IN THIS ARTICLE

1. 01Direction one: run Delyt from Claude
2. 02Direction two: give Delyt's own agents your tools
3. 03Isn't letting an AI operate your support platform risky?
4. 04Why almost no CX tool can do this
5. 05Who this is for

Every customer-support tool assumes you'll live in its dashboard. You log in, you click through queues, you configure rules in their menus. The software is the place you go.

We think that's backwards. The fastest-growing teams already live somewhere else: in an AI assistant. So we built Delyt so the whole platform — support, sales, and social — can be operated from Claude or GPT, with no UI at all.

The mechanism is the Model Context Protocol (MCP), the open standard for connecting AI clients to tools. Most companies that mention MCP implement one side of it. Delyt implements both directions — and that turns out to matter a lot.

Direction one: run Delyt from Claude

Connect Delyt to Claude (or any MCP client) and you get 50+ governed tools to operate your workspace by conversation instead of clicks. A few things you can say:

• "Show me every case waiting on us for more than 24 hours, grouped by channel." → Claude queries your cases and returns a table.
• "Draft a reply to case #4821, grounded in our refund policy." → it pulls the case, searches your knowledge base, and writes a grounded draft.
• "Create a flow that routes refunds over $200 to a human." → it builds the automation in a deactivated state for your review.
• "Schedule a month of content across Instagram, LinkedIn, and X." → it composes and schedules through the publishing suite, with a dry-run option first.
• "Pull last week's resolution KPIs and break them down by agent." → it runs the report and renders it.

Same platform, no dashboard. You operate where you already work.

Direction two: give Delyt's own agents your tools

The other direction is just as important. Delyt's agents can reach out to your stack mid-conversation through MCP — check inventory, look up a record, book a meeting, trigger a workflow. Native connectors for n8n and Airtable, Shopify and Stripe integrations in preview, plus Slack, HubSpot, Notion, Calendly, or any other MCP server. The agent doesn't just answer; it acts in the systems where the work actually happens.

Isn't letting an AI operate your support platform risky?

That's the right question, and it's why the how matters as much as the what. Every MCP connection is OAuth-scoped — you grant each one exactly what it needs (read-only reporting, or full case handling, but no config access, for example). Everything is rate-limited and logged. Write actions support a dry-run mode, so you can preview before anything happens. And Delyt's agents run on a governed engine that's hard-stopped from inventing answers and follows your playbooks step by step.

Why almost no CX tool can do this

Two reasons. First, it's architectural: legacy platforms were built UI-first, with the database locked behind their dashboard — exposing it cleanly as an open protocol is a rebuild, not a feature flag. Second, it's commercial: their business model is per-seat, so a headless, no-UI way to operate the product cuts against how they charge. We don't have either constraint. Delyt is flat-priced and built open from the core — which is exactly why we can hand you the keys.

Who this is for

If your team already runs on Claude or GPT, or you're the kind of operator who'd rather type an instruction than hunt through five menus, this is for you. It's also for builders: the MCP surface means you can compose Delyt with the rest of your AI tooling instead of treating it as a walled garden.

You can try all of it on a 14-day free trial — bring your own AI key at zero markup, or use ours. The dashboard is there if you want it. But you won't always need it.