Integrations
:::summarybox learn Why integrations exist: data in for context, actions out for delivery How to connect one (OAuth or API token, scoped per brand) What you unlock once connected: Automations, Actions, Agent, Sites, Crawler tracking, Exports The full directory, organised by purpose Security, permissions, and what breaks when you disconnect
How to connect an integration
The flow is the same shape almost everywhere.
- Open Integrations from the sidebar to reach the hub at
/integrate, or use the inline Connect button near a feature's empty state. - Pick the service. Most use OAuth: a one-click handoff to the provider's consent screen, then back to Trakkr. A handful (Cloudflare API token, Looker Studio personal API key, raw webhooks) need a credential you paste in.
- Pick the destination. Slack asks for a channel, Sheets asks for a spreadsheet, Linear asks for a team. This is where the "what gets used" decision lives, and you can change it later.
- Test it. Most cards have a Test connection button that sends a sample event so you can confirm the wiring before you depend on it.
Connections are scoped per brand, not per workspace. If you manage three brands, each gets its own Slack workspace pick, its own Linear team, its own Sheets file. That's deliberate, because your acme.com brand and your beta.com brand almost always belong in different channels.
What integrations unlock
The connection itself is plumbing. The value is the features that depend on it.
| Feature | What it does | The integration it needs |
|---|---|---|
| Automations | Triggers like visibility drops or new citations fire actions like a Slack post, a Linear issue, or a Sheets row | Whatever sink you want the action to land in |
| Actions | Every finding has a "Push to ..." button that files it as a ticket, message, or page | Linear, Jira, Slack, Notion, GitHub, Trello, Asana |
| Agent | In chat, the Agent prepares a suggestion and acts only after you confirm. When standing automations are available, they can deliver findings and make only the reversible site changes you granted for that work area. | The delivery or site connection that action needs |
| Sites | Trakkr publishes content and supported fixes through WordPress, Webflow, Shopify, or GitHub. Write and revert coverage differs by platform and fix type; see Sites. | A supported Sites connection |
| AI Crawler Tracking | Joins matching server requests to normalized pages. User-agent labels are useful signatures, not proof of the operator behind a request. Coverage and delay depend on the source. | Cloudflare, Vercel, Netlify, or your edge of choice |
| Live Visitors | Tells you how many ChatGPT, Perplexity, Gemini, or Claude referrals turned into pageviews and conversions | Google Analytics 4 |
| Exports | Live data into Sheets or Looker Studio, scheduled or on demand | Sheets, Looker Studio (Scale plan), CSV |
If a feature looks empty when you open it, the usual reason is the integration it depends on hasn't been connected yet. The empty state will tell you which.
The directory
Grouped by purpose, not provider. A service that does more than one thing (WordPress, for example) is listed where it spends most of its time.
Inbound, bring data in
| Integration | What it gives Trakkr |
|---|---|
| Google Analytics 4 | Traffic and conversions attributed to ChatGPT, Perplexity, Gemini, and Claude referrals |
| Cloudflare | Server-side AI crawler hits via the Cloudflare Analytics API. Works on every Cloudflare plan |
| Vercel | Real-time crawler events via Log Drains. Vercel Pro or Enterprise |
| Netlify | Real-time crawler events via an Edge Function |
| WordPress | Crawler events from the Trakkr WordPress plugin, plus read access for fix publishing |
| Webflow | Crawler tracking via Cloudflare proxy. Its native Sites connection handles CMS content and supported SEO fields, but not schema markup |
| Shopify | Crawler tracking via Cloudflare proxy. Also has a native content adapter (Sites connection) for pages, blog articles, and SEO metafields |
| HubSpot, Squarespace, Wix, Framer, Ghost | Crawler tracking only, routed through Cloudflare. There is no native content adapter, so Trakkr doesn't write content back to these CMSes yet |
| Next.js, Node/Express, Nginx, AWS CloudFront, Akamai, Fastly | Self-hosted webhooks that POST crawler hits from your own edge |
Outbound, push actions out
| Integration | What Trakkr can send to it |
|---|---|
| Slack | Automation alerts, approval requests, action buttons, weekly digests |
| Microsoft Teams | Automation alerts as Adaptive Cards |
| Discord | Automation alerts to server channels |
| Linear | Open an issue from a finding, action, or workflow trigger |
| Jira | Same, into a Jira project of your choice |
| GitHub Issues | File a finding as an issue against a repo |
| Asana | Turn an action into a task in a project |
| Trello | Drop an action card on a board |
| Notion | Create a page in a database (content backlog, action log, etc.) |
| Google Sheets | Append rows on a schedule, or on every workflow event |
| Looker Studio | Live BI dashboards across 17 datasets. Scale plan; see Looker Studio |
| CSV Export | One-click downloads from any table, no setup |
| Webhooks | Send any event payload to any HTTPS endpoint with Bearer, Basic, or API Key auth |
| Zapier, Make | Use Trakkr as a source or sink in your existing automations |
| REST API | Build it yourself; see the API reference. Scale plan |
| MCP Server | Expose your Trakkr brand to ChatGPT, Claude, Cursor, Codex, or any MCP-aware AI assistant. Any paid plan |
Security and permissions
A few things worth knowing before you connect anything.
Saved secrets stay server-side. Connected-site credentials use AES-256-GCM encryption. Automation integration responses return safe connection details, not saved tokens, webhook URLs, or secret values. Secret fields inside an exact rule are write-only after you save them.
Hosted MCP data. If you connect ChatGPT, Claude, or another hosted MCP client to Trakkr, that assistant can ask Trakkr for the same categories exposed by the MCP tools: brands, prompts, visibility scores, citations, competitors, reports, crawler and Action data, workflows, webhooks, Reddit monitoring, and related account metadata. Some tools are read-only; others can start a check or change state when you explicitly ask the assistant to do that, for example setting a brand's primary location so competitor analysis uses the right market instead of defaulting toward global or US results. Hosted clients receive short-lived connector tokens, not a raw Trakkr API key. Trakkr stores redacted MCP activity logs: tool names, timing, status, and safe input and result summaries for security, support, and product improvement, but not your full assistant conversation. You can revoke access at any time from Settings → Developer or by removing the connector.
Brand-scoped access. A connection only works for the brand it was created on. Connecting Slack to Brand A doesn't give Brand B access to that Slack workspace, even inside the same Trakkr account.
Narrow scopes. We ask for the smallest permission set the integration needs. Linear asks for issue read/write, not admin. Google Sheets asks for drive.file, not full Drive. Slack asks for the channel and incoming-webhook permissions, nothing about user identity.
Reauthorisation. OAuth tokens expire. When that happens the card shows a yellow Reconnect state, and the affected features pause cleanly instead of failing silently. One click restores them; nothing has to be rebuilt.
Audit trail. Every dispatch (Slack post, Linear issue, Sheets append) writes to the Actions audit log, with who triggered it, what payload was sent, and what the destination returned. Useful when something lands somewhere unexpected.
Common questions
Do I have to connect anything to use Trakkr?
No. The platform stands up without any integrations: prompts, scores, research, citations, competitors, and the dashboard all work in isolation. Integrations only unlock the features that need outside systems (Automations, Sites publishing, AI Crawler Tracking, Live Visitors). Add them as you need them.
My tool isn't on the list. What do I do?
Three paths. Webhooks sends arbitrary JSON to any HTTPS endpoint, which covers most internal systems. Zapier or Make bridges Trakkr to thousands of apps without code. MCP Server exposes your Trakkr data to any AI assistant that speaks Model Context Protocol. Beyond that, the REST API is the build-it-yourself route.
Why do CMSes like Shopify and Wix route through Cloudflare?
Those platforms don't expose server-side request logs we can read directly. The workaround is to front the site with Cloudflare so AI crawler hits land in Cloudflare's analytics, which we can read. The CMS keeps serving the site, and nothing in the publishing workflow changes.
Can the same Slack workspace be connected to multiple brands?
Yes. Each brand has its own destination picker, so you can send Brand A alerts to #brand-a-monitoring and Brand B alerts to #brand-b-monitoring from one Slack install. The OAuth grant is per brand, but the underlying workspace can be shared.
What happens to data already sent if I disconnect?
It stays where it is. Past Slack messages, Linear issues, Sheets rows, cached Looker Studio data: all of it persists in the destination. Disconnecting only stops new dispatches.
Where do I see what an integration actually sent?
Each integration card has a View Logs action that shows recent sends with a timestamp, status code, request details, and response. Automation checks have their own log under Automations. The Actions audit dashboard brings the history together across brands.