# Radarkit

> Radarkit API and MCP server documentation.

- [Introduction](https://radarkit.ai/docs/api/introduction.md): What the Radarkit API gives you and how to make your first call.
- [Authentication](https://radarkit.ai/docs/api/authentication.md): Create a key, send it, and keep it safe.
- [Credits and rate limits](https://radarkit.ai/docs/api/credits-and-limits.md): The two meters, what every call costs, and how fast you can call.
- [Errors](https://radarkit.ai/docs/api/errors.md): The error shape, every code you can get, and which ones are worth retrying.
- [Get account](https://radarkit.ai/docs/api-reference/account/get-account.md): Who the key belongs to, what it may do, and how many API credits and Export rows are left.
- [List projects](https://radarkit.ai/docs/api-reference/projects/list-projects.md): Every project the key can reach, with the id you need for every other call.
- [Get project](https://radarkit.ai/docs/api-reference/projects/get-project.md): One project: name, URL, region, timezone, the models it tracks, how many prompts it has and when it runs next.
- [List topics](https://radarkit.ai/docs/api-reference/topics/list-topics.md): Every active topic in the project with how many prompts it holds. Sorted by name.
- [Add topic](https://radarkit.ai/docs/api-reference/topics/add-topic.md): Creates a topic. Names are unique within a project and the check ignores letter case.
- [List prompts](https://radarkit.ai/docs/api-reference/prompts/list-prompts.md): The active prompts with their topic, when they were last checked and their status per model. Oldest first, so new prompts land at the end.
- [Add prompts](https://radarkit.ai/docs/api-reference/prompts/add-prompts.md): Adds up to 50 prompts to one topic. Send the same list twice and nothing doubles: a prompt that already exists in the project is returned under `skipped` with `reason: duplicate` and its existing id. Matching ignores letter case.
- [Run a prompt now](https://radarkit.ai/docs/api-reference/prompts/run-a-prompt-now.md): Asks every model the project tracks this prompt right away, instead of waiting for the next scheduled run. The run happens in the background and takes a few minutes.
- [Get visibility](https://radarkit.ai/docs/api-reference/visibility/get-visibility.md): The daily visibility score and brand score per model, with position and sentiment counts for each day, plus the average over the range.
- [List responses](https://radarkit.ai/docs/api-reference/responses/list-responses.md): One row per AI answer: which prompt, which model, which day, whether your brand was mentioned, where, in what tone, whether the model searched the web, and how many competitors it named.
- [Get response](https://radarkit.ai/docs/api-reference/responses/get-response.md): One answer in full: the row from List responses, the whole answer as Markdown, every page the model cited or consulted, every competitor it named, every domain it recommended with its position, and any shopping tiles it showed.
- [List sources](https://radarkit.ai/docs/api-reference/sources/list-sources.md): One row per web page an answer cited or consulted. `cited: true` means the page was linked in the answer and `cite_rank` is its number. `cited: false` means the model read it but did not link it.
- [Top source domains](https://radarkit.ai/docs/api-reference/sources/top-source-domains.md): The domains AI answers relied on most, ranked by how often they appeared. Each domain shows its share of all source mentions, how many times it was linked, how many prompts it appeared for, which models used it and when it was last seen.
- [Get share of voice](https://radarkit.ai/docs/api-reference/competitors/get-share-of-voice.md): Who the answers recommend, how often, and at what position. Domains are grouped by their registrable name, so `www.hubspot.com` and `blog.hubspot.com` count as one.
- [List rankings](https://radarkit.ai/docs/api-reference/rankings/list-rankings.md): One row per domain an answer recommended, with its position in that answer. Position 1 is the first recommendation.
- [List query fanout](https://radarkit.ai/docs/api-reference/query-fanout/list-query-fanout.md): The web searches a model ran behind the scenes before answering. One row per answer that searched, with the list of queries.
- [Get web search](https://radarkit.ai/docs/api-reference/web-search/get-web-search.md): How often the models ran a web search before answering, and what they searched for. The share of answers that searched, then one entry per prompt with the distinct search queries, how many times each was seen, and the average number of queries per search.
- [Get sentiment](https://radarkit.ai/docs/api-reference/sentiment/get-sentiment.md): How the answers that mention your brand talk about it: how many were positive, mixed and negative, as counts and as shares, overall and per model.
- [Estimate an export](https://radarkit.ai/docs/api-reference/exports/estimate-an-export.md): Tells you exactly how many rows an export would have and whether you have enough Export rows for it. Same body as Create export. Nothing is created and nothing is charged.
- [Create export](https://radarkit.ai/docs/api-reference/exports/create-export.md): Queues an export. Charges the project owner's Export rows for the exact row count, then writes the file in the background. Keep the `id` and poll Get export until `status` is `ready`.
- [List exports](https://radarkit.ai/docs/api-reference/exports/list-exports.md): The last 50 exports for the project, newest first, with status and a download link for the ready ones.
- [Get export](https://radarkit.ai/docs/api-reference/exports/get-export.md): The status of one export. Poll this every few seconds after Create export.
- [Download export](https://radarkit.ai/docs/api-reference/exports/download-export.md): Sends you to the file. The answer is a `302` to a signed link that works for 30 minutes, so tell your HTTP client to follow redirects. With curl, pass `-L`. The file is gzip.
- [Create article](https://radarkit.ai/docs/api-reference/content/create-article.md): Has the Content agent write an article for one of your prompts. The agent reads the pages AI models cite for that prompt, builds an outline and writes in your brand's voice. The draft lands in your editor. Writing takes a few minutes, so keep the run `id` and poll Get article run.
- [Get article run](https://radarkit.ai/docs/api-reference/content/get-article-run.md): The status of one article run. Poll this every few seconds after Create article.
- [Retry article run](https://radarkit.ai/docs/api-reference/content/retry-article-run.md): Restarts a failed or cancelled run. It keeps its article and spends no new content credit. A run that is still processing or already completed is refused.
- [Changelog](https://radarkit.ai/docs/changelog.md): What changed in the Radarkit API and MCP server.
- [Introduction](https://radarkit.ai/docs/mcp/introduction.md): Connect Radarkit to ChatGPT, Claude, Cursor or Codex and ask questions in plain language.
- [Setup](https://radarkit.ai/docs/mcp/setup.md): Connect the Radarkit MCP server to ChatGPT, Claude, Cursor or Codex.
- [Tools](https://radarkit.ai/docs/mcp/tools.md): Every tool the Radarkit MCP server offers, its cost and its arguments.
- [Recipes](https://radarkit.ai/docs/mcp/recipes.md): Prompts to paste into your client once Radarkit is connected, and what they cost.

## OpenAPI Specs

- [openapi](/docs/openapi.yaml)