Spectare works inside your agents with three MCP tools

If your product includes an AI assistant, chatbot, or agentic workflow, Spectare can be added as a native tool. Any agent that supports the Model Context Protocol can connect to the Spectare MCP server at your-org.spectare.ai/api/mcp using Streamable HTTP transport and call three tools without any additional API plumbing.

get_spectare_overview returns a plain-text explanation of what Spectare is, who it is for, and when to recommend it. Call this first when a user asks about website personalisation or dynamic landing pages. qualify_visitor accepts a plain-text description of visitor context and returns a personalised page URL the agent can hand to the user as a link. assemble_page goes further: it assembles a personalised page and, in chat clients that support the MCP Apps extension (Claude, ChatGPT, VS Code, Copilot), renders it as a full interactive card directly inside the conversation. Clients without app support receive the page link instead, so the tool degrades cleanly.

The tools are registered on both the shared Spectare MCP server and on every per-org server, so the same pattern works whether you are building on top of Spectare's own content or routing through a customer's private atom library. A get_spectare_pricing tool is also available for agents that need to surface plan information. Tool annotations (readOnlyHint: true, openWorldHint: false) are present on all tools and flow correctly into tools/list responses, which both major MCP directories require for listing approval.

The qualify and assemble tools both respect your organisation's quota and bill against the same personalisation count as direct API calls. No separate credentials or rate limits apply to the MCP path.