Competitor monitoring agent
Public-page changes, with the exact evidence and a response topic to review.
Competitor pricing and launches change quietly. A remembered announcement is not evidence of what changed.
Built for read-only checks; verify coverage and register the daily task before treating it as running · Competitor sites · Risk: Low. Everything on this page is public; there is nothing to sign in for.
Tasks and outputs
What it takes on
Check public pages
Respects robots.txt, limits concurrency and waits between requests to the same domain.
Compare evidence
First fetch establishes a baseline. Failed fetches never replace a successful snapshot.
Propose responses
Calls Claude only when changes exist. Requires a verbatim diff quote and a strategy quote for each topic.
Who it is for: Small AEO teams that need evidence-backed response topics, not another social feed.
How we use it: Built for our own monitoring workflow. The initial target list covers verified public pricing and blog pages for Profound and Peec. Changelog extraction and the remaining vendor URLs still need verification. Scheduling is not enabled by this website.
Evidence before interpretation
01No login, captcha bypass or social monitoring.
02No change means no model call.
03New targets establish a baseline, not a launch announcement.
04No automatic publishing or gap-analysis edits.
05The public log updates only after a site deployment.
The rules the agent’s output is written to, from the file named above. This site prints no example output; what the agent has shipped is in its run log.
Workflow
How a run goes, step by step
Pick a step to see what goes in, what comes out, and where a person is involved.
Step 1 of 4
Fetch
Fetches only verified targets, respecting robots.txt and per-domain spacing.
targets.json
Normalized public content or a fetch failure
Verify URLs and extraction rules.
Step 2 of 4
Compare
Compares the latest successful snapshot before today. Keeps known entry IDs to avoid re-reporting old posts.
Successful snapshots
Changes with before and after evidence
Review extraction noise.
Step 3 of 4
Interpret
One tool-free Claude session reads the diff and strategy. Quotation validation rejects unsupported output.
Diff and strategy
Interpretations and optional response topics
Approve or drop topics.
Step 4 of 4
Record
Writes summary.md and appends deduplicated topics. The website reads the summary at build time.
Validated evidence
Run record and topic pool
Review, commit and deploy records. No automatic Git publishing is enabled.
Before you start
What you’ll need to bring
Company material
A positioning strategy to quote when proposing response topics.
Source material
Verified public target URLs. Missing vendors remain an explicit coverage gap.
Platform account
No competitor account. A local Claude Code login is required only for analysis.
Run environment
Node.js 22 or later. Optional Playwright. Windows Task Scheduler for daily unattended checks.
Run log
Recent runs, as they happened
The full log is public and read-only. It holds every run on record, including the ones that stopped to wait for a person.
View the full run logGuides and files
Set it up yourself
The setup guide, and the files the agent is built from. You run it in your own environment, with your own accounts.
- 01
Runtime
- Node.js 22 or later, Git and Claude Code for changed-page analysis.
- 02
Browser setup
- HTTP is the default. Optional rendered targets use the existing Playwright installation with a fresh, unsigned-in browser.
- 03
Account setup
- No competitor accounts or credentials. Claude Code must already be signed in locally.
- 04
Configuration
- Verify target URLs, extraction rules and robots.txt before adding them. The vendor roster comes from AI Visibility.
- 05
First run
- Run into a temporary output directory, inspect the baseline and failures, then register the daily task on the scheduling machine.
targets.json
Config · JSON. Verified targets and outstanding coverage.
fetch.mjs · diff.mjs · summarize.mjs
Scripts · Node. Public fetching, deterministic comparison and constrained interpretation.
run.mjs
Runner · Node. Writes dated records, local snapshots and deduplicated topics.
run-competitor-watch.ps1
Scheduler · PowerShell. Daily execution with locking and local transcripts. No automatic commit or deploy.
Not packaged for public download yet. The guide on the left describes what each one does, step by step, so you can rebuild the setup.
Good to know
Where it stops and asks for you
- A target fails or is blocked. Records the failure without replacing its snapshot. Three consecutive days trigger a maintenance note.
- Analysis fails validation. Keeps the diff and reports that analysis needs attention. Does not advance changed targets.
- Nothing changed. Writes a successful unchanged record without calling Claude.
Risk notice
What to watch for
- Page extraction can miss content or include layout noise. Review evidence before making a competitive claim.
- Failed pages keep the previous successful snapshot. They never count as a change.
- Topic suggestions are not articles, endorsements or automatic publishing decisions.
FAQ
Questions about this agent
Does opening this page start a check?
No. This page reads repository records at build time. A script on the scheduling machine runs the check.
Would you rather we ran this agent for you? There is a managed service for that.
Talk to usOther agents