Your AI visibility data, outside the dashboard
The dashboard lets you read evidence one brand at a time. Some work needs the data elsewhere: a client report in your own format, a spreadsheet that combines several brands, a script that adds questions in bulk or an AI agent that can answer how a brand did last week. SearchSeal offers two ways in. The REST API returns JSON over HTTPS. AI agents and coding tools can call the MCP server’s tools for the same data.
Every plan includes both at no extra cost. Both use the same account, the same tokens and the same request limits.
Tokens with read or write access
In account settings, create a personal access token to send as a Bearer token. Each token is either read or write. A read token can fetch data and cannot change anything. A write token can also add questions, run page checks and change the status of fixes. Use a separate token per script or agent to revoke one without breaking the rest.
Create and revoke tokens only in the app. The API cannot issue new tokens, change your plan or touch billing. The authentication guide covers token scopes and how to rotate them.
What the REST API covers
Version 1 of the API lives under /api/v1 and is brand scoped. Read endpoints return your account and plan, the AI tools on your plan, your brands, tracked rivals, questions, saved answers with their details, visibility reports by brand, domain and URL, the content of returned source pages, search performance, fixes, before and after comparisons, page checks and articles.
Write endpoints add up to 50 questions per call, archive or restore a question, accept, dismiss or complete a fix, run a page check on your own domain and draft an article. Page checks and articles use the same monthly allowance as the dashboard. Responses show when they were generated and, where relevant, the date the data runs through. Code generators can use an OpenAPI description. See the API reference for every endpoint, filter and field.
Connect an AI agent with MCP
An AI agent uses the Model Context Protocol to call server tools. SearchSeal runs an MCP server at https://searchseal.com/api/v1/mcp over Streamable HTTP. Point an MCP client at that address with your token, and it can list brands, read answers and reports, review fixes and before and after results, and read page checks and articles.
With a write token, the agent can also add questions, archive a question, update a fix, run a page check and start an article. The server tells clients to ask you before adding questions or changing a fix, and marks those tools as changes rather than reads. Four saved workflows start the agent off: a weekly pulse, top fixes, whether a fix worked and a scorecard by AI tool.
The MCP setup guide has configuration for a command line client, an HTTP client and a local bridge for desktop apps.
Limits and what you cannot do
Requests are limited per account, across REST, MCP and every token, with an hourly and a 30 day limit. Response headers show the remaining allowance. The current numbers are in the API reference.
The API does not create, edit or delete brands, and it never deletes data. Archive or restore questions while keeping their history. Page checks and articles work only on your brand's own domain. MCP sign-in uses tokens only, with no OAuth flow. SearchSeal marks answer text and source page content as untrusted for agents because both come from third parties. Treat them as data to inspect, not instructions to follow.
Saved answers are samples from specific dates, tools and locations. Moving them into another report does not change what they can show: they are evidence for a decision, not a ranking.