CitedSpy for Activepieces
Track how your brand appears in AI answer engines - ChatGPT, Perplexity, Gemini, Copilot and Google AI Mode - and drive automations from it.
CitedSpy runs your tracked prompts against the real consumer UIs of those engines and records what they said: whether your brand was mentioned, in what position, with what sentiment, and which sources they cited. This piece connects that data to Activepieces.
Install
The piece is not in the official Activepieces list, so install it by name:
Settings → My Pieces → Install Piece → @citedspy/piece-citedspy
Connect
Authentication is a CitedSpy API key.
- In CitedSpy, go to Settings → API keys
- Click New key, name it "Activepieces", and copy the value
- Paste it when Activepieces asks for a connection
The key starts with cspy_ and is shown only once. API keys are included from
the Growth plan up; on Solo or a trial the connection is refused with a
message explaining why.
Your connection is labelled with the workspace name and plan, so several CitedSpy workspaces are easy to tell apart.
Triggers
| Trigger | Fires |
|---|---|
| New Alert Triggered | Instantly, when a CitedSpy alert fires: sentiment turns negative, a competitor overtakes you, or a new citation source appears |
| New Run Completed | When a prompt finishes running on an AI engine |
| New Brand Mention | When your brand or a tracked competitor is named in an answer |
New Alert Triggered is a real webhook. The other two poll.
Every trigger can be scoped to one brand. New Brand Mention also filters by sentiment and by whether the mention is of your brand or a competitor - setting sentiment to Negative is the most useful configuration, because it tells you when an engine says something bad about you without notifying you about every neutral mention.
Actions
| Action | Does |
|---|---|
| Create Prompt | Adds a tracking prompt to one of your brands |
| Trigger Prompt Run | Queues a prompt to run on the AI engines now |
| Custom API Call | Any other CitedSpy v1 endpoint |
Trigger Prompt Run costs money. Each queued run is a real query against an AI engine and counts toward your usage. Do not put it on a frequent schedule.
Custom API Call reaches everything this piece does not wrap yet - dashboard
metrics, the full report for a prompt, citations and recommendations. Give it a
path such as /dashboard/YOUR_BRAND_ID; the base URL and your key are added for
you. See the API reference.
Examples
Negative mentions to Slack - New Brand Mention with Sentiment Negative → Slack Send Message, including the prompt, the engine and the snippet.
A run history in Sheets - New Run Completed → Google Sheets Insert Row, mapping the engine, prompt, answer preview and cost.
Alerts become issues - New Alert Triggered → Linear or Jira Create Issue, using the alert type, brand and dashboard link.
Notes
- Only enabled engines are offered. The engine picker lists engines switched on for your workspace, because filtering a flow to a disabled engine would mean it never fires.
- Answers are shortened in lists. New Run Completed returns the first 500
characters plus the true length. Use Custom API Call against
/runs/YOUR_RUN_IDfor the complete text. - Polling reads the newest 200 records per check. A one-off burst larger than that can overflow a single check; normal scheduled refreshes are well inside it.
Links
License
MIT