This guide describes what each category of endpoint enables and the shape of data it returns. Response fields shown in examples are illustrative — your own data, competitors, and citations will reflect your workspace.
Why the API instead of the dashboard or MCP
The dashboard is for a person looking something up. MCP is for a person asking a question through an assistant. The API is for a system that needs to act without anyone in the loop — running on its own schedule, reacting to events, or living inside a product other people use. Reach for it when:- Something needs to happen on a recurring schedule (a nightly report, a weekly digest) with no manual step.
- You’re building a feature into your own product that surfaces KIME data to your own users or customers.
- You want to react automatically to a change (a sentiment drop, a new competitor ad) rather than someone noticing it.
- You’re integrating with another system (a PM tool, a CRM, a data warehouse, a CMS) where the sync itself should be code, not a conversation.
Dashboards and white-label reporting
Pull visibility, share-of-voice, sentiment, and a composite performance score into your own UI instead of sending people to log into KIME directly. A performance-style endpoint returns one composite scorecard, well suited to a KPI tile:Scheduled reporting and digests
The same read endpoints, called on a schedule (a cron job or a scheduled function) instead of on page load, formatted into whatever channel your team actually reads: a Slack or Teams message, an email digest, a generated PDF, or a row appended to a spreadsheet. A recurring job can pull performance scores across every tracked workspace or brand, rank them, and post a summary to an internal channel on a fixed cadence — fully unattended, with no one needing to remember to check a dashboard.Threshold-based alerting
Poll the analytics endpoints on a schedule, diff against the previous period, and trigger a notification when something crosses a threshold you define. Useful signals to alert on include:- A meaningful period-over-period sentiment score change
- A competitor’s share-of-voice overtaking yours on a tracked query set
- Sponsored ads beginning to appear in your category’s AI answers where they weren’t before, via the ads-share endpoints
Two-way sync with your project management tool
The Actions Center backlog is fully readable and writable via the API, not just for display purposes. Each task carries a task type (content creation, content update, “get mentioned in this article,” “pitch for exclusive content,” “add a trust signal,” and others), a priority, a status that moves through a defined lifecycle (suggested, to-do, in progress, in review, done or rejected), and a full brief — source URLs with citation counts, which of your tracked prompts it would help win, and either a content brief or extracted outreach contact details (name, email, phone) for mention/PR-type tasks.The full task lifecycle is scriptable end to end: create a task, comment on it, fetch it individually or in bulk by ID, and update its status — the same operations your dashboard performs are available as direct calls.
Automated content pipeline: gap to draft to publish
Chain the Actions Center to the content generator to your actual publishing target. A completed content-generation job’s export includes full SEO metadata (title, slug, meta description, primary and secondary keywords), Open Graph and social card data, a ready-to-use structured-data block (including a populated FAQ schema where relevant), and the complete article body — available in multiple formats, including a WordPress-compatible import format alongside plain JSON, Markdown, and HTML. A pipeline can watch for an approved content task, poll until the linked generation job completes, pull the export in whichever format your CMS accepts, and import it automatically — going from “here’s a content gap” to “here’s a published or ready-to-publish page” with no manual writing step and no manual export/import click. Gate the final publish step behind a human approval if you want a review checkpoint without giving up the rest of the automation.Bulk prompt and tracking provisioning
Create up to several hundred prompts in a single call, each tagged with the right location and category. This is the backbone of two common operational patterns: an agency onboarding flow that generates a templated prompt set (branded, non-branded, and comparison queries) programmatically from a client’s product and competitor list and submits it in one call instead of a team member typing dozens of prompts into a UI by hand; and a new-market or new-product launch checklist that provisions tracking automatically as part of a broader launch automation, polling for completion before the first report is due.AI-crawler monitoring in CI/CD
A website-health-style endpoint runs a real crawl and scoresrobots.txt and structural accessibility individually for major AI crawlers (OpenAI’s, Anthropic’s, Perplexity’s, and others) — not a single pass/fail, but a per-bot, per-category breakdown.
Wire this into a deploy pipeline, or a scheduled job that runs independently of deploys, so a change that accidentally blocks an AI crawler fails the build or opens a ticket automatically — the AI-search equivalent of a broken-link checker, and something a classic SEO crawler doesn’t check at all. Because the response includes category-level scores and per-page detail, it can also feed a “GEO readiness” tile alongside existing performance and accessibility metrics in an internal engineering dashboard.
Blending AI-visibility data into your own data warehouse
Pull the analytics endpoints into your data warehouse (Snowflake, BigQuery, Redshift, or similar) on a schedule and join against your own CRM, product, or revenue data. This answers questions no single tool can answer alone — for example, whether a rise in AI share-of-voice for a given product line correlates with pipeline or signups some weeks later — by putting AI-visibility metrics next to the rest of the company’s data instead of leaving them in a separate silo only accessible through a dashboard.Outreach and PR CRM sync
Mention and exclusive-content-pitch tasks come back with the citing page’s author name, email, phone, or contact-page URL already extracted as structured data, not something you have to look up manually. Feed these directly into an outreach tool or CRM as ready-made contact records, so a PR or partnerships team’s pipeline starts with “here’s who to contact and why it matters” instead of a manual research step for every opportunity.Multi-workspace and portfolio orchestration
Every operation above is scoped to a single workspace, but nothing stops you from looping across many. Agencies, holding companies, and platforms managing several brands can build a thin orchestration layer that enumerates workspaces, runs the same pull or sync logic against each, and aggregates the results — a single script maintaining dozens of client integrations instead of one-off manual setups per client.Embedding KIME data as a feature in your own product
If you’re building a product that already serves a marketing, SEO, or brand-management audience, the same read endpoints can back a native “AI visibility” feature inside your own application rather than sending users to a separate tool — your product handles the UI and your own auth, and calls out to the API server-side to populate it.Notes for building on the API
- Pagination: list endpoints return
items(or a similarly named results field),totalCount,limit, andoffset— page by incrementingoffsetbylimituntil it reachestotalCount. Defaults and maximums vary by endpoint. - Large payloads: citation analysis, the advertiser directory (which can include embedded image data for creative previews), and the AI website health check (full per-page crawl detail) can return large responses on an active, heavily-tracked workspace — request narrow date ranges, use available filters, and handle pagination in code rather than assuming a single response is complete.
- Sentiment scale: a 0–100 scale, where a score above the midpoint is positive, below it is negative, and exactly at the midpoint is neutral.
- Engine filtering: most analytics endpoints accept a filter for a specific AI engine or search surface, letting you build an engine-by-engine breakdown instead of only a blended number — useful if your audience or strategy differs meaningfully by platform.
- Location format: prompt locations use standard two-letter country codes (or a recognized country name normalized to one) — broad regions aren’t accepted as a location value.
- Idempotency and retries: for anything that writes (creating prompts, creating or updating tasks), design your automation to handle retries safely — check current state before re-submitting rather than assuming a failed-looking response means nothing happened.