Playbooks · Skill & signal
A saved prompt that pulls job postings, ad activity and company posts for your target accounts through the hosted MCP server, and writes a short digest. RichAPI has no scheduler and no alerting product — the cron is yours. Every endpoint in this prompt bills per result returned, on any 2xx, so cap the account list before you cap anything else.
Last updated September 22, 2026
Pattern: account list → one MCP prompt → linkedin_job_search + linkedin_ad_search + linkedin_company_posts → digest you read on Monday
Credit math: Worked example at stated counts, one weekly run over 25 accounts: linkedin_job_search returning 6 postings per account = 150 results at 0.1 = 15 credits. linkedin_ad_search returning 4 ads for 10 of them = 40 results at 0.2 = 8 credits. linkedin_company_posts returning 5 posts per account = 125 results at 0.2 = 25 credits. Roughly 48 credits per run, about 208 a month at weekly cadence, before any contact lookup. 0.1 credits per linkedin_job_search result, 0.2 credits per linkedin_ad_search result, 0.2 credits per linkedin_company_posts result
ROI math: The cost is small enough that the real constraint is your attention, not your balance. Twenty-five accounts produces a digest a human reads in five minutes; two hundred produces a document nobody opens. Convert credits at your tier rate on /pricing.
Owned internally by: Founder or AE running a named-account list
Best for: Anyone working 20 to 50 named accounts who wants a reason to reach out this week rather than a queue of undated leads
# The MCP server is hosted. Point your client at it and the endpoints
# arrive as tools — no local server, no wrapper code.
#
# host: mcp.richapi.ai
# auth: your RichAPI API key
#
# The scheduling half is yours. A cron entry that opens the prompt:
# 0 8 * * MON /usr/local/bin/run-weekly-digest.sh
Here are my 25 target accounts, as LinkedIn company URLs: [paste list].
For each account, using RichAPI:
1. Search job postings for that company. Flag any posting for a role in
revenue operations, data engineering or demand generation.
2. Check for recent LinkedIn ad activity.
3. Pull the company's five most recent posts.
Then write me a digest, ordered by how strong the signal is. For each account
worth acting on, give me one sentence on what changed and one sentence I
could actually open an email with. Skip accounts with nothing new — do not
pad the list. Tell me the total result count you retrieved.
Sign up at app.richapi.ai and copy your key.
25 free credits, no card.
Connect the hosted MCP server at mcp.richapi.ai in Claude and authenticate with your key.
The endpoints show up as tools.
Save the prompt above with your real account list pasted in.
That file is the whole skill.
Run it by hand for two weeks.
If you are not reading the output, do not automate it — you will have built a scheduled way to ignore something.
Once it earns its place, schedule it yourself: a cron entry, a GitHub Action, a Shortcut, whatever you already run. We do not provide one.
Cap the account list and the result counts in the prompt.
Per-result billing means the list length is the bill.
There is no digest feature to enable. What you are building is a text file with your account list in it and three instructions, pointed at the hosted MCP server, run on a schedule you own. That sounds unimpressive until you compare it to the alternative, which is checking 25 companies by hand and quietly checking 4. **[Get 25 free credits — no card](https://app.richapi.ai)**
Worth being blunt, because the word "weekly" in the title implies something we do not sell. RichAPI has no cron, no watch lists, no triggers, no webhooks that fire when a company posts a job, no Slack notifications and no email digests. We are an API and a hosted MCP server. Anything recurring happens because you scheduled it. That is fine, because you already own a scheduler. A crontab line. A GitHub Action on a `schedule` trigger. An n8n or Make workflow if you prefer clicking. A calendar reminder if you are being honest about how this usually goes. Any of them work, and none of them require us to build an alerting product we would then have to run well.
**Job postings.** The most legible signal on this list. A company hiring three data engineers has a data problem with a budget attached, and the posting is dated, public and specific about the stack. **Ad activity.** Evidence that money is moving through a marketing function right now. Thin on its own, useful stacked with the other two. **Company posts.** What the company says it cares about this quarter, in language you can quote back. This is where the opening line comes from. None of it is intent data. Nobody raised a hand, we do not sell an intent feed, and a job posting is a fact about a hiring plan rather than a fact about a buying cycle. The value is that all three are dated — you are writing about something that happened this month, which is most of what separates a reply from a delete.
| Call | Count per weekly run | Rate | Credits | |---|---|---|---| | `linkedin_job_search` | 25 accounts × 6 postings = 150 | 0.1 / result | 15 | | `linkedin_ad_search` | 10 accounts × 4 ads = 40 | 0.2 / result | 8 | | `linkedin_company_posts` | 25 accounts × 5 posts = 125 | 0.2 / result | 25 | | Per run | | | **48** | Rates used above: 0.1 credits per `linkedin_job_search` result, 0.2 credits per `linkedin_ad_search` result, 0.2 credits per `linkedin_company_posts` result. All three bill per result on any 2xx. They are not multi-provider waterfall endpoints, so nothing here is free on an empty return, though per-result billing does mean an account with no activity is close to it. Convert at your tier rate on [the pricing page](/pricing). The number that moves this bill is the account count, and it scales linearly. Two hundred accounts is roughly eight times 48 credits a week, which is still small, and a digest nobody finishes reading.
**Automating before reading.** The digest that runs every Monday into a folder you never open costs credits and produces nothing. Run it by hand twice first. If you skip week two, delete the file. **Asking the model to decide what matters, then ignoring what it decides.** The prompt above tells it to skip accounts with nothing new. Keep that instruction. The moment you let it pad to a fixed length you get five paragraphs of "no significant changes this week" and you stop reading by week four. **Letting the model invent the signal.** It is summarising tool output, and it is capable of smoothing a thin result into a confident sentence. Ask for the result count, as the prompt does, and spot-check an account against the raw output monthly. **Treating an MCP run as a batch pipeline.** Conversational runs over 25 accounts are the right shape for this. A thousand accounts belongs in a script hitting the HTTP API directly, where you control retries and can log every call. **Signals with no owner.** A digest is not a workflow. Decide on day one who writes the email when an account lights up, because "we all saw it" means nobody did.
The individual signals each have a fuller page: [hiring signal to outbound](/use-cases/hiring-signal-to-outbound) for the jobs branch, [Meta ads to an account list](/use-cases/meta-ads-to-account-list) and [Google Ads transparency to outbound](/use-cases/google-ads-transparency-to-outbound) for advertising evidence, and [the use-case index](/use-cases) for the funding and hiring variants, which are also crons you run yourself. [Agent-driven account research](/use-cases/agent-driven-account-research) goes deeper on one account instead of shallow across many. Setup for the MCP server is on [the Claude integration page](/integrations/claude), and [the pricing page](/pricing) has the credit tiers.
**Does RichAPI run this on a schedule for me?** No. No scheduler, no alerting, no notifications. You own the cron and we are not planning to take it from you. **Can I get a webhook when a target company posts a job?** No. Polling on your schedule is the mechanism. That is also why the account count matters more than anything else in the prompt. **Which MCP host do I connect to?** `mcp.richapi.ai`, authenticated with your API key. **Does this work in something other than Claude?** Yes. Any MCP client. [ChatGPT](/integrations/chatgpt) and the rest connect to the same host with the same key. **Do these endpoints cost nothing when an account has no activity?** They bill per result, so no results means close to no cost. That is arithmetic, not the waterfall guarantee. Only `email_finder`, `email_verifier` and `phone_finder` return `billed: false` on a soft fail. **Should I run this over my whole CRM?** No. This is a named-account play. A digest is only useful at a length a person will read on a Monday morning. **Do I need a credit card?** No. 25 free credits on signup.
Paste five accounts into the prompt and run it once. If the output gives you one email you would actually send, scale the list to 25 and put it in your crontab. **[Get 25 free credits](https://app.richapi.ai)**