Connect Manifold and Context.dev to Claude: one prompt pulls a competitor's rankings, ads, social and site, then finds three gaps for this week. Free to start.
Swipe or tap Next · checked on 2026-10-13
Install Claude Code first if you don't have it (step 1 under “Connect both to Claude Code” below), and make your free Context.dev account at context.dev (no card). Manifold makes your account the first time you sign in.
Then run these lines in your terminal: they make a folder, plug both MCPs in, and start Claude. Inside Claude Code, type /mcp, pick manifold and sign in in your browser, then do the same for context. Then paste the prompt. It walks you through the rest and stops before anything spends a credit.
Using the Claude app instead? Do “Or: connect both to the Claude app” below, then paste the same prompt into a new chat with both switched on.
Make the folder, add both MCPs, start Claude
cd ~
mkdir competitor-check
cd competitor-check
claude mcp add --transport http --scope user manifold https://mcp.manifoldmcp.com/mcp
claude mcp add --transport http --scope user context https://mcp.context.dev/mcp
claudecd ~
mkdir competitor-check
cd competitor-check
claude mcp add --transport http --scope user manifold https://mcp.manifoldmcp.com/mcp
claude mcp add --transport http --scope user context https://mcp.context.dev/mcp
claudeThe setup prompt (copy all of it) · lines 1-1 of 40. Copy all copies the whole prompt; read it in full on the page.
About me: I run a home-service business and I'm not technical, so use plain words and short sentences.About me: I run a home-service business and I'm not technical, so use plain words and short sentences.
Goal: research one competitor with public data (their Google rankings, the searches they show up for and I don't, their ads, their Facebook and Instagram, what ChatGPT says when people ask for what we sell, and their website) and give me three gaps I can go after this week. We'll use two MCPs (cables that let you use other apps): Manifold for the marketing data and Context.dev to read websites. Follow this guide; it's the source of truth for every step and every price:
https://receiptsgroup.com/guides/competitor-teardown
My website: <yoursite.com>
What I sell, and where: <your service> in <your towns or area>
The competitor: <theirsite.com>
Their Facebook page: <page name or link, or "find it">
Their Instagram: <@handle, or "skip">
Steps:
1. Check you have both sets of tools: Manifold's (names starting seo_, aeo_, ads_, facebook_, instagram_) and Context.dev's (web-scrape, web-map, get-usage). If one is missing, tell me to type /mcp, pick it and sign in in my browser (slash commands are mine to type, not yours). Then explain in two sentences what an MCP is and what a credit is.
2. If you're Claude Code: write CLAUDE.md in this folder from the guide's box, with my details filled in, and show it to me before you save it. In the Claude app, skip this and follow the rules below for the whole chat.
3. The free first test: run Context.dev's get-usage and tell me my balance and when it refills. Then run Manifold's seo_get_page on the competitor's home page and tell me its title, its main headings and what it seems to sell. Both cost 0 credits.
4. Write three questions a customer would ask ChatGPT when they need what I sell where I sell it. Show them to me and wait. In Claude Code, once I approve them, add them to CLAUDE.md under "My ChatGPT questions".
5. Price the whole teardown before anything runs. The calls: Manifold seo_get_domain_overview on their site (5 credits); seo_get_keyword_gap with my site as the target and theirs as the competitor, 100 rows (10); facebook_get_profile on their page (1); ads_get_advertiser_ads on facebook with that page's ad_library_id and active_only on, one page (1); ads_search_advertisers on google with their business name (1); ads_get_advertiser_ads on google with that advertiser id, no details, one page (1); instagram_get_profile (1); aeo_run_ai_answers with my three questions on chatgpt only, with my business and theirs as the brands (2 a question). Context.dev: web-scrape their home page as markdown in compact mode (1); web-map their site (1); web-map my site (1). Dry run every Manifold call (dry runs are free), add Context.dev's prices, and show me a table: call, what it's for, credits, with a total for each service. Wait for my yes.
6. Run it. Manifold calls with max_credits set to the dry-run price. Context.dev calls one at a time. After each call, tell me what it actually cost.
7. Show me what you found, in short sections: Google (how strong their site is, their top pages, the 20 biggest searches they show up for and I don't), their website (what they sell, the pages they have that I don't), their ads (offers, hooks, how long they've run), their social (followers, how often they post) and ChatGPT ("when I asked today", who got named).
8. Three gaps for this week: things I could start in a week, like a page, a listing fix or an offer to test. Most useful first. For each: what to do, the result it comes from, and what the data can't tell us. In Claude Code, save it as a new file with today's date in the name; in the Claude app, just show it.
Rules for every call:
- Price it first and wait for my yes. Manifold and Context.dev charge the moment a tool runs. They don't ask first, so you have to.
- Public company data only. Never use Manifold's leads_ tools or Context.dev's people-enrich, and never look up a person or an email. If a result includes a person's name, email or phone number, leave it out of everything.
- Don't use web-crawl, web-answers, submit-batch or create-monitor unless I ask: they cost more, and a monitor keeps spending credits every time it runs.
- One Context.dev call at a time. The free plan runs one at a time.
- If a call fails or comes back empty (NoData), tell me and stop. Don't run it again: an empty Manifold result still costs a credit.
- Use only what the results show. Never invent a ranking, an ad, a number or a business name. It only sees what's public, so never guess their sales, their leads or their close rate.
- About ChatGPT, say "when I asked today", never "ChatGPT recommends". One run is a snapshot.
- Never say a gap will get me ranked or get me calls. It's something to try, not a promise.
- Don't copy the competitor's words, ads or titles into my plan.
- Never ask me for, print or save a key, token or password.
House rules for this setup:
- Go one step at a time and wait for me after each step.
- If you use a technical word, explain it in plain English the first time.
- Ask me before you assume anything about my business, my accounts or my setup.
- Stop and check with me before anything that costs money or spends credits.
- Stop and check with me before anything that sends, posts or publishes to real people.
- Stop and check with me before you delete or overwrite anything.This is the manual way: every step, every command and every price. Claude follows it too.
Run the same prompt on the same competitor next month and compare. One run is a snapshot; the change is the part worth reading.
yoursite.com.Everything here is stuff anybody could look up: Google data, the public ad libraries, public profiles, public web pages. Their close rate, their leads, their call recordings, how their reps run an appointment? No clue, and it shouldn't have one.
The Google numbers are estimates from a data provider (DataForSEO), the ad libraries publish spend as ranges, and ChatGPT's answer can change from run to run. So treat the three gaps as good questions to work on, not facts about their business.
This is for planning your own work. Don't post the teardown, don't put their name in your ads, and never publish a claim about a named competitor off the back of it. One run on one day doesn't prove anything about them, and it's their name on the line. And it's company data only: no person lookups, no emails. The prompt keeps it that way.
Skip this if claude --version already prints a version. You need a paid Claude plan: Pro, Max, Team or Enterprise. The free plan doesn't include Claude Code. The first time you run claude, it sends you to your browser to log in.
Install (Mac, Linux, WSL)
curl -fsSL https://claude.ai/install.sh | bash
# open a NEW terminal window, then:
claude --versionInstall (Windows PowerShell)
irm https://claude.ai/install.ps1 | iexClick Create a free account. The free plan is 1,000 credits a month, one request at a time, and no card. You don't need to do anything like this for Manifold: your first sign-in makes the account, with 250 free credits and no card.
Your rules file and your gap lists land in here, not all over your computer. If the folder's already there, mkdir complains; ignore it.
Make the folder and go into it
cd ~
mkdir competitor-check
cd competitor-checkcompetitor-check folder. Not inside Claude Code.Two lines, one per MCP. --scope user adds each one for every folder on your computer. That's how Context.dev's own instructions do it, and Manifold's say to add --scope user if you want it everywhere. Already set one of these up from my Manifold guide or my Context.dev guide? Then Claude Code says that server already exists. Fine: skip that line.
Add Manifold and Context.dev
claude mcp add --transport http --scope user manifold https://mcp.manifoldmcp.com/mcp
claude mcp add --transport http --scope user context https://mcp.context.dev/mcpclaude first, then /mcp.Pick manifold in the list and follow the steps in your browser: sign in to Manifold with Google or email, and approve Claude Code. Then open /mcp again, pick context, sign in to Context.dev and pick your organization on the consent screen. Rather skip the menu? At the plain terminal prompt, claude mcp login manifold and claude mcp login context run the same sign-ins.
Type this in Claude Code
/mcpContext MCP in your account (a key is a password for apps; you never need to see this one). Claude Code keeps both sign-ins, and claude mcp remove deletes them. The first time Claude uses a Manifold tool, Claude Code asks your permission.At the plain terminal prompt, claude mcp list should show manifold and context, both ✔ Connected. If one says ! Needs authentication, do its sign-in again. Or just ask Claude: “What Manifold and Context.dev tools do you have?” You should see Manifold names starting with seo_, aeo_, ads_, facebook_ and instagram_, and Context.dev's web-scrape, web-map and get-usage among the rest.
Check from the terminal
claude mcp listHere's the part nobody mentions: both of these charge the moment a tool runs. Neither one asks first. So the asking is Claude's job, and this file is where you tell it. Claude Code reads it at the start of every session. It's an instruction, not a lock, though. The lock on the Manifold side is max_credits: Manifold's own server refuses any call priced over it. On the Context.dev side, watch the first few calls and say no if Claude skips the price.
CLAUDE.md · Part 1 of 2 · Copy copies all 30 lines
# Competitor teardown
This folder is for researching one competitor with public data, using Manifold MCP (marketing data, read-only) and Context.dev (reads websites), and turning it into three gaps for this week.
## My business
- Website: <yoursite.com>
- What I sell, and where: <your service> in <your towns or area>
- The competitor: <theirsite.com>
- Their Facebook page: <page name or link>
- Their Instagram: <@handle>
## My ChatGPT questions (the same ones every month)
1. <question 1>
2. <question 2>
3. <question 3># Competitor teardown
This folder is for researching one competitor with public data, using Manifold MCP (marketing data, read-only) and Context.dev (reads websites), and turning it into three gaps for this week.
## My business
- Website: <yoursite.com>
- What I sell, and where: <your service> in <your towns or area>
- The competitor: <theirsite.com>
- Their Facebook page: <page name or link>
- Their Instagram: <@handle>
## My ChatGPT questions (the same ones every month)
1. <question 1>
2. <question 2>
3. <question 3>
## Every call
1. Price it first: dry run every Manifold call (dry_run: true, free), add Context.dev's price (web-scrape 1, web-map 1), show me the total and wait for my yes.
2. Run Manifold calls with max_credits set to the dry-run price. Run Context.dev calls one at a time.
3. Tell me what each call cost (Manifold: credits_charged; Context.dev: credits_consumed and credits_remaining).
4. ChatGPT checks: chatgpt only, unless I ask. Say "when I asked on <date>", never "ChatGPT recommends".
## Never
- Never look up a person or an email: no leads_ tools, no people-enrich. Leave names, emails and phone numbers out of everything.
- Never use web-crawl, web-answers, submit-batch or create-monitor unless I ask.
- Never invent a ranking, an ad, a number or a business name. Only what the results show. Never guess their sales, leads or close rate.
- Never rerun a call that failed or came back empty (NoData). Stop and tell me.
- Never say a gap will get me ranked or get me calls.
- Never copy the competitor's words, ads or titles into my plan.
- Never ask me for, print or save a key, token or password.CLAUDE.md · Part 2 of 2 · Copy copies all 30 lines
## Every call
1. Price it first: dry run every Manifold call (dry_run: true, free), add Context.dev's price (web-scrape 1, web-map 1), show me the total and wait for my yes.
2. Run Manifold calls with max_credits set to the dry-run price. Run Context.dev calls one at a time.
3. Tell me what each call cost (Manifold: credits_charged; Context.dev: credits_consumed and credits_remaining).
4. ChatGPT checks: chatgpt only, unless I ask. Say "when I asked on <date>", never "ChatGPT recommends".
## Never
- Never look up a person or an email: no leads_ tools, no people-enrich. Leave names, emails and phone numbers out of everything.
- Never use web-crawl, web-answers, submit-batch or create-monitor unless I ask.
- Never invent a ranking, an ad, a number or a business name. Only what the results show. Never guess their sales, leads or close rate.
- Never rerun a call that failed or came back empty (NoData). Stop and tell me.
- Never say a gap will get me ranked or get me calls.
- Never copy the competitor's words, ads or titles into my plan.
- Never ask me for, print or save a key, token or password.Works in Claude on the web and in the desktop app. You're adding two custom connectors, and Claude's free plan allows one, so plan on Pro or up for both. (Context.dev's install page says it's also in Claude's connector directory; Claude's help page doesn't say whether a directory connector counts toward that free limit.) No folder and no CLAUDE.md here, so the rules ride along in the setup prompt.
Claude's help center says Customize, then Connectors. Manifold's own page says Settings, then Connectors. Use whichever your app shows you.
Click + Add, then Add custom connector. Name it Manifold, paste the address below and click Continue. Claude shows the sign-in settings it detected; click Continue again.
Manifold's connector address
https://mcp.manifoldmcp.com/mcpUnder Authentication, pick Sign in now. Under OAuth client, pick Use Claude's published identity (Recommended). Leave Request headers empty: that's for API keys, and you're signing in through the browser instead. Click Add. Then sign in to Manifold (Google or email) and approve Claude. If no window opens, find Manifold in your connectors and click Connect.
Back in Connectors: + Add, Add custom connector, name it Context, paste the address below, and make the same choices: Sign in now, Claude's published identity, no request headers, Add. Then sign in to Context.dev and pick your organization. (Context.dev's own page shortens all this to: +, Add custom connector, the address, Connect.)
Context.dev's connector address
https://mcp.context.dev/mcpStart a new chat, click the + button at the lower left, then Connectors, and make sure Manifold and Context are both on. Then paste the setup prompt from “Start here”. Claude asks before it uses a tool. Claude's help center says to pick Allow always only for tools you trust to run unsupervised. On the paid calls I'd keep saying yes one at a time.
An Owner adds them once for everyone: Organization settings, Connectors, Add, hover over Custom, pick Web, then the same names and addresses. Everyone else goes to Customize, Connectors, finds each one and connects.
Three calls that cost nothing: Context.dev's balance check, Manifold's free snapshot of one page, and a dry run of the biggest Manifold call in the teardown. If all three come back, both cables work. Paste this into Claude Code (or your Claude chat with both on).
Prompt: the free first test
Use Context.dev's get-usage (it's free) and tell me my credit balance and when it refills.
Then use Manifold's seo_get_page (it's free) on <https://theirsite.com>. Tell me the page title, the main headings and, in one sentence, what they seem to sell.
Then do a dry run only (dry_run: true) of Manifold's seo_get_keyword_gap with <yoursite.com> as the target and <theirsite.com> as the competitor, 100 rows. Tell me what it would cost. Don't run it.Use Context.dev's get-usage (it's free) and tell me my credit balance and when it refills.
Then use Manifold's seo_get_page (it's free) on <https://theirsite.com>. Tell me the page title, the main headings and, in one sentence, what they seem to sell.
Then do a dry run only (dry_run: true) of Manifold's seo_get_keyword_gap with <yoursite.com> as the target and <theirsite.com> as the competitor, 100 rows. Tell me what it would cost. Don't run it.Good: your Context.dev balance and refill date, their page title and headings, a price of 10 credits for the keyword gap, and nothing charged. (Manifold's answers say what they cost in credits_charged; Context.dev's say credits_consumed.) Bad: Claude says it has no Manifold or Context.dev tools, or the answer says it needs a sign-in. See “When it breaks”.
Here's the prompt, with the guardrails written in. Swap in your site, what you sell, the competitor and your three questions. Claude prices the whole thing first and waits for your yes.
Prompt: the teardown · lines 1-4 of 11. Copy all copies the whole prompt; read it in full on the page.
Research this competitor and give me three gaps I can go after this week.
Me: <yoursite.com>, <your service> in <your towns or area>.
Them: <theirsite.com>. Facebook page: <name or link, or find it>. Instagram: <@handle, or skip>.Research this competitor and give me three gaps I can go after this week.
Me: <yoursite.com>, <your service> in <your towns or area>.
Them: <theirsite.com>. Facebook page: <name or link, or find it>. Instagram: <@handle, or skip>.
Manifold: their domain overview; the searches they rank for and I don't (seo_get_keyword_gap, me as the target, 100 rows); their Facebook page, then their active Facebook ads using the page's ad_library_id; their Google ads (find the advertiser first, no ad text); their Instagram profile; and these three questions on ChatGPT only, looking for my business and theirs: <question 1>, <question 2>, <question 3>.
Context.dev: their home page as text (web-scrape, compact), then web-map their site and mine, one at a time.
First, dry run every Manifold call, add Context.dev's prices, show me the total and wait for my yes. Then run them, Manifold with max_credits at the dry-run price, and tell me what each one cost.
Then show me what you found in short sections (Google, their website, ads, social, ChatGPT), and the three gaps: what to do, the result it comes from, and what the data can't tell us. Public company data only. Nothing invented, no person lookups, don't copy their words. Say "when I asked today" about ChatGPT.seo_get_domain_overviewseo_get_keyword_gapfacebook_get_profileads_get_advertiser_adsads_search_advertisersads_get_advertiser_adsinstagram_get_profileaeo_run_ai_answersweb-scrapeweb-mapPrices from Manifold's and Context.dev's docs, checked 2026-10-13. That's about 26 Manifold credits and 3 Context.dev credits. The dry run gives you the real Manifold total before anything runs, and an answer that comes from Manifold's cache costs less.
How the gaps come together: the keyword gap says which searches they show up for and you don't, and the page they show up with. Your own site map says whether you even have a page for that town or service. Put the two side by side and you get the one from the video: a town where they rank and you've got no page.
Their ads show what they're pushing, their home page shows how they say it, and the ChatGPT questions show who got named when you asked. Claude has to tie every gap to one of those results. If it can't point to one, it's not a gap, it's a guess.
Before you build anything, see where you stand for the search behind a gap. One search, one location.
Prompt: where do I rank for this search?
Use Manifold's seo_get_position. Where does <yoursite.com> rank for "<a search from the gap list>"? Try the location "<your town or state>". Dry run first and wait for my yes. If I'm not in the top 100, just say so.No Manifold or Context.dev call, just Claude working from what's already in the chat.
Prompt: the page plan
Take gap 1 and write me a plan for the page: the town, the service, the page address following my own site's pattern, the headings, and the questions a customer in that town would want answered. My words, not theirs. No numbers I didn't give you, and no promise that it will rank.Take gap 1 and write me a plan for the page: the town, the service, the page address following my own site's pattern, the headings, and the questions a customer in that town would want answered. My words, not theirs. No numbers I didn't give you, and no promise that it will rank.The teardown skips the ad text to keep it cheap. If their Google ads look worth a closer read, this adds it, at 25 credits a page instead of 1.
Prompt: the Google ad text
Use Manifold's ads_get_advertiser_ads on google for the same advertiser, with details on, one page. Dry run first and wait for my yes. Then sum up the offers and the wording they repeat. Don't copy it into my plan.The same teardown again, about the same price. The change is what tells you something.
Prompt: what changed
Run the same teardown on <theirsite.com> with the same three ChatGPT questions from CLAUDE.md. Price it first and wait for my yes. Then compare with the file from <date>: new searches they show up for, ads that started or stopped, pages they added, and which of my three gaps I've closed.Run the same teardown on <theirsite.com> with the same three ChatGPT questions from CLAUDE.md. Price it first and wait for my yes. Then compare with the file from <date>: new searches they show up for, ads that started or stopped, pages they added, and which of my three gaps I've closed.The keyword gap is the workhorse: 10 credits for the 100 biggest searches they show up for and you don't. The pricier calls are easy to reach for and rarely needed here: crawling their whole site on Context.dev is 1 credit a page, and asking Claude and Perplexity too is 6 credits a question each, against 2 for ChatGPT. Start small, and add more once you know what you're looking for.
Checked 2026-10-13 on manifoldmcp.com/pricing, context.dev/pricing and claude.com/pricing. Manifold's launch post said 500 free credits; its pricing page says 250, so plan on 250. Manifold's plan credits reset every month, and there's no overage: when credits run out, calls stop. Context.dev also lists bigger plans (Growth, Scale, Enterprise), says yearly billing saves two months, and its docs say unused credits on paid plans generally don't roll over. Prices change: check before you buy.
get-usagedry_run: trueseo_get_pageget_taskseo_get_domain_overviewseo_get_keyword_gapseo_get_positionfacebook_get_profile, instagram_get_profileads_get_advertiser_adsads_search_advertisersaeo_run_ai_answersNoData)web-scrapeweb-mapweb-crawlFrom Manifold's credits page and tool docs and Context.dev's credits page, checked 2026-10-13. Manifold's cached answers cost a quarter of the live price (at least 1 credit), and nothing if your workspace already paid for them. AI-answer runs are never cached.
Rough math for one teardown (from the price lists, not a bill): Manifold about 26 credits (domain overview 5, keyword gap 10, Facebook page 1, Facebook ads 1, Google advertiser 1, Google ads 1, Instagram 1, three ChatGPT questions 6).
Context.dev 3 (their home page 1, two site maps 2). So Manifold's free 250 covers about nine teardowns, and Context.dev's free 1,000 a month won't be the thing you run out of. Add the 6-credit Google check for each gap you want to test.
How Manifold charges: it holds the most a call could cost, then charges for what actually came back, never more than the estimate. If your balance is too low, the call fails and costs nothing. An empty answer (NoData) still costs 1, which is why the prompt says don't rerun it.
command not found: claude, or 'claude' is not recognizedOpen a new terminal window after installing. Still missing? The install folder isn't on your PATH yet: Claude Code's install troubleshooting page has the fix for your system.
You added it in an earlier guide. Nothing's wrong: skip that line and go on to the sign-in.
claude mcp list says ! Needs authentication, or Claude says it has no Manifold or Context.dev toolsYou added it but didn't sign in. Start claude, type /mcp, pick the one that's missing and finish in your browser. Or run claude mcp login manifold or claude mcp login context at the plain terminal prompt.
OAuth token is invalid or expired (Manifold)Sign in again. In Claude Code: /mcp, pick manifold, then Re-authenticate. In the Claude app, remove the connector and add it again (that's how Claude's help center says to change one).
Context.dev's fix: clear the old sign-in and connect again. In Claude Code, claude mcp logout context, then claude mcp login context. In the Claude app, remove the connector and add it again.
Name the tool and ask for a live lookup: “Use Context.dev's web-scrape on theirsite.com.” That's Context.dev's own advice. Also check you haven't said no when Claude asked to use the tool.
InsufficientCredits (Manifold)Your balance is below that call's price. Nothing was charged. The free 250 come once, so it's a Pro plan, a top-up or next month's renewal. Claude should stop and show you the link the error carries.
BudgetExceeded (Manifold)The call was priced over your max_credits cap, so Manifold refused it and charged nothing. That's the cap doing its job. Ask for fewer rows, or raise the cap once you've said yes to the price.
NoData (Manifold)The data provider had nothing for that input, and that still costs 1 credit. Don't run the same thing again. Check the spelling of the domain or the page name. A small competitor can just have no Google data, no Instagram or no ads.
The ad library works off their Facebook page. Have Claude get the page first (facebook_get_profile): it says whether the page is running ads now, and gives the ad_library_id the ads tool takes. If it's not running ads, there's nothing to find, and that's an answer too.
Google's ad library is keyed by an advertiser id, and one business can have one entry per region. Have Claude find the advertiser first (ads_search_advertisers, 1 credit) and use the id it returns.
Normal. The ad libraries publish spend and impressions as ranges, so that's all anybody gets. Reach numbers only show up for Facebook ads that ran in the EU or the UK.
The location defaults to the United States. The tools take a location by name, so ask Claude to dry-run it with your state or city. If the tool rejects the name, that costs nothing; fall back to the country. Town names often show up inside the searches themselves anyway.
It runs in the background. Claude gets a task number and checks back with get_task, which is free. If Claude gives up, ask it to check the task again. Results stay there for 30 days.
Normal. Manifold's own docs say the answers are live and change from run to run. Ask the same three questions every month and read the trend, not one run.
429 from Context.dev: too many requestsThe free plan runs one request at a time, and Claude sent two at once. Nothing was charged. Ask it to run the Context.dev calls one after the other.
WEBSITE_BLOCKED on the web-scrapeTheir site answered with an anti-bot page, a CAPTCHA or a login screen. Don't run the same scrape again unchanged. The site map and all the Manifold data still work without it.
truncated)Context.dev caps each answer at 40,000 characters. Ask for less: the compact mode on a scrape, or fewer pages on a map. Running the same thing again gets the same cut and costs again.
Claude Code warns when a tool sends back more than 10,000 tokens (chunks of text) and caps it at 25,000 by default. Ask for fewer rows: the 20 biggest searches, not 1,000.
401 USAGE_EXCEEDED (Context.dev)You're out of Context.dev credits. The free plan has no overage, so calls stop until your monthly refill. Ask Claude to run get-usage (free) to see when that is, or move to a paid plan. Don't retry: it won't change.
401 DISABLED (Context.dev)Your key's been switched off. Context.dev's docs say free accounts can have all their keys turned off after repeated failed requests, which is why the prompt tells Claude to stop on errors. Stop retrying, and check your email from Context.dev or contact their support.
Ask Claude to show you the exact result behind it: the search, the page address, the ad. If it can't point to one, cross it off. You know your towns and your customers better than any data provider does.
Different company, same name. This one is manifoldmcp.com, and the address you add is https://mcp.manifoldmcp.com/mcp.
Claude Code: claude mcp remove manifold and claude mcp remove context, which also delete the sign-ins they stored. Claude app: Customize, Connectors, then Remove (or the three dots) next to each. Then, as Context.dev's install page says, disable the Context MCP key on your API keys page (context.dev/dashboard/api-keys). Manifold says it deletes your account data within 30 days of you deleting the workspace or closing the account.
Manifold logs every call with the domain or keyword you asked about, and passes the inputs to its data providers: DataForSEO for the Google and ChatGPT tools, and a social-data vendor for the ad and social tools. Its privacy page lists them all. Context.dev's sign-in key spends your own organization's credits. Neither one needs anything private from you, so keep customer names, phone numbers and anything private out of your prompts.
What each page backs up is listed at the end of the full guide.
One prompt, two cables, three gaps, and it turns out your competitor's been doing your homework.
Would you rather have someone build those pages on a site you own? That's our Custom SEO.
Want help setting this up? Receipts Group builds these systems.
Open the full guide as a page (every file in full, plus the table of contents).