Competitive Intel Agent: Track every competitor. Catch every move. Turn it into action
Competitor monitoring for marketing teams. Track every rival in your market, get a cited weekly report, and turn any signal into content from the same workspace.
Quick answer
The Competitive Intel Agent runs once a week against each competitor in your Magi market.
It researches six things per competitor: launches, positioning, pricing, customer sentiment, GTM moves, and press.
It reads your own knowledge base too: sales calls, decks, win-loss notes, Slack.
Each signal carries a source. External ones cite a URL. Internal ones cite the file and the excerpt.
You approve the findings, or turn review off and let it publish.
The report includes your angle and an "if they say, we say" talk track per competitor. Any signal becomes a draft in one click.
Competitor monitoring tools stop at the alert
Crayon, Klue, and Kompyte watch rival websites and send alerts. The alerts go to Slack or a battlecard. A product marketer reads them, and most of the time nothing happens next. The signal was three tools away from the draft.
We built the Competitive Intel Agent so the signal and the draft live in one place. It watches the competitors you already listed in your Magi market. It ranks what they did by how much it matters to your buyer, writes the report, and hands any signal to the Content Agent.
Where it sits in BrandOS. The Research Agent fills your knowledge base from every source you point it at. The Competitive Intel Agent reads that same base plus the open web. It does one job with it: competitor monitoring with a point of view on how to respond.
What the agent researches every week
For each tracked competitor, six questions, scoped to the last seven days:
Product moves. Launches, feature changes, deprecations.
Positioning and messaging. Homepage and category changes, new claims.
Pricing. Plan changes, new tiers, packaging.
Customer sentiment. Reviews, community threads, churn signals.
GTM signals. Hiring, partnerships, funding, executive moves.
Press and analyst coverage. With the tone of each piece.
Each finding becomes a signal. A signal has a headline, two or three sentences of detail, a type, a relevance rating (high, medium, or low), and a source. Each research field also keeps its basis: the URLs, the excerpts, and a confidence rating for the claim.
How it works
External research. One deep research pass per competitor. The agent is briefed with your brand name, personas, market topics, and the competitor's site, blog, and LinkedIn page. It is told to lead with what changed and to cite a URL for every claim. Uncited claims are dropped.
Internal research. The agent reads the summary of each item in your knowledge base. It pulls out anything that names a tracked competitor: a customer call where they came up, a lost-deal note, a partner deck. Each internal signal keeps the filename and the exact excerpt.
Merge. External and internal signals are combined into one research report per competitor. This step is plain code. Nothing gets added or reworded.
Review. The run pauses. You see each competitor's summary and signals. Approve, edit, or remove any of them. The citation evidence is locked; only the reading of it can change. If you would rather not review, turn on auto-publish and skip this step.
Synthesis. The only step that gets your full brand voice, product positioning, and ICP pain points. For each competitor it writes your angle, an "if they say, we say" talk track, and the week's dominant themes and sales implications. This runs on Claude Opus 5.
Report. The Content Agent writes the competitive intelligence report from your template. It appears as a draft in Build, with each signal still linked to its source.
What you see in Magi
Open Workforce > Competitive Intel Agent.
The agent home shows four numbers: signals that need your attention, competitors tracked, reports this period, and weeks running. Below that, the run calendar marks each scan as a square. Click a day to see what was detected.
Active Competitors lists each tracked competitor with its signal count for the period, split by high, medium, and low. It also shows whether activity is up or down against last month. Click one to open its latest signals: date, headline, detail, relevance, and a link to the source.
The right panel logs Actions, Outputs, and Runs. Each action reads like "Review Knowledge, Sep 4, 2026: 1 report, 1 draft," so you can trace what a run produced and where it went.
How to configure the Competitive Intel Agent
Click Configure on the agent page. Seven settings.
Competitors. By default the agent tracks all competitors in your markets. Remove any, add one by name, or pull in another market's list.
Intel brief. Tell the agent what matters this quarter. "We launch pricing tiers in October. Watch pricing pages." The brief shapes what ranks high.
Research sources. Two toggles. Deep Research runs the external pass. User Uploaded reads the knowledge docs you add here.
Whitelist and blacklist. Domains the agent may cite, and ones it must ignore. Social posts, job boards, app stores, and SEO comparison sites are blocked at the citation gate before you add anything.
Integrations. Slack and Google Drive as internal sources.
Run schedule. Weekly by default. Pick the day, time, and timezone. The panel shows the next run.
Notifications. Get told when a run finishes and a report is ready for review, so nobody has to check the agent page.
How teams run it
What a dedicated competitive intelligence platform costs you
Klue and Crayon run $15,000 to $40,000 a year. Kompyte starts lower. All three are built to deliver alerts and battlecards to sales. Reviews of the mid-market tools name the same two problems: stale data, and more alerts than anyone reads. The usual fix is a person who curates.
Magi puts the curation in the pipeline. Signals are ranked by fit to your ICP. The report arrives with your angle in it. And the draft is one click away, in the workspace where your team already writes. That is competitive marketing intelligence, built for the team that has to respond.
Two things we are still fixing
Each signal is cited, but the agent cannot yet confirm that each date is right. Some sites stamp today's date on old pages. A check against the page's own dateline exists and is being tuned before it ships on by default.
The report covers the competitors you configured. If research turns up a new entrant, it will not get its own row until you add it.
See the report on your own competitor list
Book a demo and we will run it against your market.
Already on Magi? Login and open Workforce > Competitive Intel Agent.
Frequently asked questions
What is competitive intelligence? Competitive intelligence means tracking what competitors do and using it to make better decisions. That covers launches, pricing, positioning, and hiring.
How do you gather competitive intelligence? Pick the competitors, decide which sources you trust, check them on a schedule, and record each finding with its source. In Magi the Competitive Intel Agent does this weekly and writes the report.
What is in a competitive intelligence report from Magi? A ranked table of signals per competitor, each with a headline, detail, relevance, and source. Then your angle, an "if they say, we say" talk track, the week's themes, and sales implications.
How often does the Competitive Intel Agent run? Weekly by default. You set the day, time, and timezone, or trigger a run by hand.
Does the agent need human review? Review is on by default. You can turn it off so reports publish automatically.
Where does the agent get its information? From web research on the competitors you track, limited to sources you whitelist. And from your own knowledge base: call recordings, uploaded documents, and Slack.
