What the server does
#The Searchly MCP server is at https://app.trysearchly.com/api/mcp. It uses stateless Streamable HTTP and the same workspace API key as the REST API, sent as a bearer token. There is no OAuth sign-in.
Every tool is an API operation with the same name. Your client sees only the tools its key may call, and each tool says whether it only reads data. Set up a client, browse the tools or follow a recipe.
Read your project data
#Discover the available tools, then call listProjects to find your project ID. Tools that act on an existing project require its explicit path.projectId; your browser's selected project does not affect MCP requests.
Tool names match the API's operation IDs. The server provides 141 JSON tools before permission filtering, covering projects, prompts, reports, answers, AI shopping, Content Studio, integrations, and Agent Analytics. Discovery lists only tools for which your key has every required permission.
Arguments separate path identifiers, query filters, and write body values. Use typed numbers for limits and string arrays for supported list filters. The visibility example uses inclusive UTC dates; report windows are limited to 366 days, and omitted dates select the last seven days.
Read the API reference for each operation's inputs and required permissions. The tool's discovery schema describes its MCP arguments. For Webflow disconnectIntegration, pass the top-level ifMatch value from current connection metadata; it supplies the API's If-Match header.
Results and limits
#Tool results contain readable JSON text and the same object in structuredContent. Successful data uses the API's data and meta.requestId envelope. Application failures set isError: true and preserve safe error codes, conflict details, and request IDs.
The getWorkspace result's limits object can include crawler_access, api_access, and mcp_access. These fields are reserved and false on every plan, and a result may omit them. They do not show whether API, MCP, or crawler access is enabled for your workspace.
_meta["searchly/httpStatus"] identifies the operation's HTTP status. Throttles also provide _meta["searchly/retryAfter"] in seconds. Calls retain the API's 60 requests per minute per key and 300 per workspace, authentication limits, and native feature limits. See errors and rate limits for retry guidance.
Use the paginated listAnswers tool to inspect answer history. Follow each collection's pagination fields; full answer history is not returned in a single tool result.
exportAnswersCsv remains a streaming REST endpoint at GET /api/v1/projects/{projectId}/answers/export and is not exposed as an MCP tool. Use the same bearer key and supported report filters through the CSV export endpoint. MCP does not buffer that download into a tool result.
Reports keep their observed evidence, availability, and measurement labels. A successful response does not prove fresh provider data or completed generation.
Client compatibility
#Use a client that accepts an API key in a custom authorization header. Clients that require OAuth sign-in cannot connect. Cookies and browser sessions do not authenticate MCP, and the older HTTP+SSE /sse transport is not provided.
The stateless endpoint receives POST requests. GET and DELETE return 405; accepted notifications return 202. Requests are limited to 1 MiB, and responses are not cached.
Connect through https://app.trysearchly.com/api/mcp. Preview and immutable deployment URLs have separate Vercel deployment protection: a Searchly API key alone cannot access them. Authorized preview testing requires Vercel access or a deployment protection bypass header in addition to the Searchly key. Keep deployment protection enabled.
Provider OAuth consent remains a separate interactive owner workflow. MCP does not sign in to Google or Webflow for you. Follow the integration workflow to connect and configure providers.