Competitor data your coding agent can call, not scrape
RivalSweeper runs a remote MCP server with 35 competitor tools: prices and discount depth, new products, live ads, promotions, homepage changes, tech stack and more, for the brands you track. Add one URL to Claude Code, Codex or Cursor, sign in with your RivalSweeper account, and your agent works from dated data instead of guessing or scraping.
Opens rivalsweeper.com, the service behind this page.
- One endpoint for every MCP client: https://mcp.rivalsweeper.com/mcp
- 33 of the 35 tools are read-only; the other 2 only file a monitoring or data request
- Included in every paid RivalSweeper plan: tool calls are counted against your plan, not billed as credits
| Tool | What it returns |
|---|---|
| get_competitor_moves_feed | What your rivals changed, newest first: prices, ads, messages, tech, shipping, hiring and more |
| get_discount_frequency_and_depth | How often and how deep each brand discounts |
| get_most_discounted_products | The products with the deepest current markdowns |
| get_newly_published_products_and_pages | New products and pages, often before they are promoted |
Tool list read from the live server on Oct 16, 2026. No sample data shown.
Anthropic introduced the Model Context Protocol on 25 November 2024 as an open standard for connecting AI assistants to the systems where data lives.
Source: anthropic.comClaude Code connects to remote MCP servers over HTTP with claude mcp add --transport http, and handles OAuth sign-in from its /mcp menu.
Source: code.claude.comCodex adds a remote Streamable HTTP server with codex mcp add <name> --url <url>, and starts an OAuth login with codex mcp login <name>.
Source: learn.chatgpt.comYour agent can write the pricing script. It can't see your competitors
Ask a coding agent to compare your prices with three rivals and it does one of two things. It writes a scraper, which breaks on the first bot check, ignores currency and returns one snapshot with no history. Or it runs a web search and summarizes whatever it finds, which is a guess dressed up as data. Neither tells you when a rival cut prices, how deep, or what they launched last week.
The data your agent needs already exists in RivalSweeper: daily reads of rivals' own storefronts, catalog changes, live ads from public ad libraries, promotions, homepage changes. MCP is how an agent reaches it. One server URL, typed tools with clear names, responses that carry when the data was captured, and access scoped to the brands your account tracks.
That turns competitor work into something you can put in a repo: a weekly brief written to markdown, a pricing check before a release, a CSV for the merch team, or a scheduled headless run that posts what changed.
Without it
- Letting the agent scrape: Quick to write, slow to keep alive. Bot checks, currency, variants and 'compare at' prices break it, and it has no…
- Web search inside the agent: Fine for a fact or a press story. It returns what pages say about a brand today, not a dated record of prices,…
- Exporting from a dashboard: Accurate, but someone has to download, paste and re-run it every week. MCP lets the agent pull the same data itself,…
With Competitor Analysis MCP
- One endpoint for every MCP client: https://mcp.rivalsweeper.com/mcp
- 33 of the 35 tools are read-only; the other 2 only file a monitoring or data request
- Included in every paid RivalSweeper plan: tool calls are counted against your plan, not billed as credits
- 35 typed tools
- What changed, in one call
From one URL to answers in your terminal
- 1
Add the server
Point your MCP client at https://mcp.rivalsweeper.com/mcp. In Claude Code that is one claude mcp add command, in Codex one codex mcp add command, in Cursor one entry in mcp.json.
- 2
Sign in once
On the first tool call the client opens the RivalSweeper sign-in page (OAuth with PKCE). Your token is scoped to your own store and the rivals your account tracks.
- 3
Ask in plain language
"What did my rivals change this week?" The agent picks the tools, such as list_my_tracked_brands and get_competitor_moves_feed, and answers from dated records.
- 4
Put it in your workflow
Have the agent write the result to a file, a CSV or a PR description, or run the same prompt headless on a schedule with claude -p or codex exec.
Features
35 typed tools
Prices and price statistics, discount frequency and depth, most-discounted products, new products and pages, catalog analysis, live ads across channels, promotions and coupons, newsletters, social, reviews, hiring, tech stack and SWOT.
What changed, in one call
get_competitor_moves_feed and get_significant_competitor_changes return what your rivals changed, newest first, so an agent starts from the delta instead of re-reading every site.
Dated, not guessed
Responses carry when the data was served and how fresh it is. A brand added in the last 24 hours says it has no data yet, rather than returning an empty guess.
Ad and deal health
find_ads_with_broken_pages_or_expired_offers checks where live Meta ads land: broken pages, sold-out products and offers past their end date, for your brand or a rival.
AI visibility signals
get_how_ai_models_describe_brand and get_ai_model_citation_rate show how large language models describe and cite brands. RivalSweeper currently samples Google Gemini and OpenAI's open-weight GPT-OSS models.
Scoped and read-only by default
33 tools only read. request_new_brand_monitoring and request_missing_data_signal file a request; nothing deletes or changes data in your account or anyone else's.
Same data as the app
The MCP reads the same records as the RivalSweeper dashboards, alerts and reports, so the number your agent quotes matches what your team sees.
The tools your agent gets
Tool names as the live server lists them. Your agent reads these names and descriptions and picks the right ones for each question.
| Tool | What it returns |
|---|---|
| get_competitor_moves_feed | What your rivals changed, newest first: prices, ads, messages, tech, shipping, hiring and more |
| get_discount_frequency_and_depth | How often and how deep each brand discounts |
| get_most_discounted_products | The products with the deepest current markdowns |
| get_newly_published_products_and_pages | New products and pages, often before they are promoted |
| get_cross_channel_ads_overview | Live ads across channels from public ad libraries |
| find_ads_with_broken_pages_or_expired_offers | Live Meta ads that land on broken pages, sold-out products or ended offers |
| get_technology_stack | The platforms and providers a brand runs on |
| get_predicted_competitor_moves | Indications of a rival's likely next move, with confidence |
Tool list read from the live server on Oct 16, 2026. No sample data shown. · See the full sample on rivalsweeper.com →
Who it's for
Developers at D2C and ecommerce brands
You already ship with Claude Code or Codex. Now the pricing script, the merch export and the release checklist can include live competitor data without a scraper to maintain.
Growth engineers and technical marketers
Turn "what are rivals running this week?" into a repeatable prompt that writes a dated markdown brief or a CSV, instead of an afternoon of tabs.
Analysts who live in the terminal
Pull discount depth, launches and ad activity for a whole competitive set into a notebook or a file, with the capture date on every number.
Teams building their own agents
No RivalSweeper plan yet? A developer API key works on the same endpoint and tools, billed as prepaid credits, with no charge when there is no data.
Works with
- Claude Code
- Codex CLI
- Cursor
- Claude (custom connector)
- Any MCP client with Streamable HTTP
- OAuth 2.1 sign-in with PKCE
- Developer API key (Bearer token)
- REST with OpenAPI (mcp.rivalsweeper.com/openapi.json)
- Official MCP Registry: com.rivalsweeper/signals
Pricing
MCP access is included in every paid RivalSweeper plan; tool calls count against a monthly plan limit, never against credits. The 14-day free trial does not include MCP, API or export. Billed monthly in USD, cancel any time, 14-day money-back on every plan. Developers without a plan can use a prepaid API key instead.
Basic
$39 per month
One developer, your store and two rivals.
- MCP included: 1,000 tool calls a month
- 2 monitored domains plus your store
- 1 user
- Up to 10,000 products per competitor
- Same data as the RivalSweeper app
- Email support within 2 business days
Pro
$199 per month
A team running agents against a full competitive set.
- MCP included: 10,000 tool calls a month
- 15 monitored domains plus your store
- 5 users
- Up to 20,000 products per competitor
- REST API, 10,000 requests a month
- Support within 1 business day
Developer API key
$0.02 per credit, prepaid
Agent builders without a RivalSweeper plan.
- Same MCP endpoint, Bearer key instead of sign-in
- Core signal = 1 credit
- No data for a domain = no charge
- Hard spend cap on every account
- No monthly fee on Pay As You Go
Compared with the alternatives
- Letting the agent scrape
- Quick to write, slow to keep alive. Bot checks, currency, variants and 'compare at' prices break it, and it has no history, so it cannot tell you what changed.
- Web search inside the agent
- Fine for a fact or a press story. It returns what pages say about a brand today, not a dated record of prices, launches and ads.
- Exporting from a dashboard
- Accurate, but someone has to download, paste and re-run it every week. MCP lets the agent pull the same data itself, on demand or on a schedule.
What changed
- Anthropic introduced the Model Context Protocol on 25 November 2024 as an open standard for connecting AI assistants to the systems where data lives. Source: anthropic.com
- Claude Code connects to remote MCP servers over HTTP with claude mcp add --transport http, and handles OAuth sign-in from its /mcp menu. Source: code.claude.com
- Codex adds a remote Streamable HTTP server with codex mcp add <name> --url <url>, and starts an OAuth login with codex mcp login <name>. Source: learn.chatgpt.com
Your first hour
If you already have a paid RivalSweeper plan, setup takes one command and one sign-in. If not, start the trial to add your rivals, then pick a plan to switch on MCP.
- Minute 1: add https://mcp.rivalsweeper.com/mcp to Claude Code, Codex or Cursor.
- Minute 2: sign in with your RivalSweeper account when the client asks.
- Minute 5: ask "What did my rivals change this week?" and check the answer against the app.
- Day 1: save the prompt as a script or a scheduled headless run that writes a dated brief.
Security and data
- Sign-in is OAuth with PKCE (S256) and automatic client registration; tokens refresh and can be revoked.
- Each token is scoped to your own store and the rivals your account tracks.
- 33 of 35 tools are read-only; the 2 request tools cannot delete or change data, and no tool takes a payment.
- Competitor data comes from public pages and public ad libraries; nothing connects to your store, ad account or analytics.
- Revoke access from your client at any time, for example Clear authentication in the Claude Code /mcp menu.
Questions
Which MCP clients does it work with?
Any client that supports remote MCP servers over Streamable HTTP. Claude Code, Codex and Cursor connect with the server URL; Claude and other assistants connect as a custom connector. The endpoint is https://mcp.rivalsweeper.com/mcp.
How does sign-in work?
OAuth with PKCE and automatic client registration. Your client opens the RivalSweeper sign-in page the first time a tool needs your account; after that the token is stored by the client and refreshed. Developers without a plan send a Bearer API key instead.
Which brands can I query?
Your own store and the rivals your RivalSweeper account tracks. To add a brand, add it in the app or ask the agent to call request_new_brand_monitoring. New brands can take up to 24 hours before data appears.
Is it included in the free trial?
No. The 14-day trial covers the app with 1 rival; MCP, API and export come with a paid plan. On Basic you get 1,000 tool calls a month, on Pro 10,000, and Enterprise is set with sales.
Can the agent change anything?
33 of the 35 tools are read-only. The other two file a request, to monitor a new brand or to add a missing data signal. Nothing deletes or edits data, and no tool takes a payment.
Which AI models are behind the AI visibility tools?
RivalSweeper currently samples Google Gemini and OpenAI's open-weight GPT-OSS models for how AI models describe and cite brands. The other tools return tracked data, not model output.
Is there a REST API too?
Yes. The same data is available over REST, described in an OpenAPI file at mcp.rivalsweeper.com/openapi.json, for pipelines that do not use an agent.
Give your agent real competitor data
Open the MCP page on RivalSweeper for the full tool list and setup, or pick a plan and add one URL to Claude Code, Codex or Cursor.
See the MCP on RivalSweeper