Skip to main content

Seven things can't do
without Calven

Connect Calven MCP to your Claude skills, and everything they write starts from trusted facts about your market, buyers and product.

Keep your Claude skills. Give them Calven.

Claude is where your team asks and creates. Calven MCP is where the facts come from.

pitch-deck/Your custom Claude skill
  • SKILL.md
  • Brand guidelines.pdf
  • Brand voice.txt
  • Pitch deck template.pptx

Calven MCPApproved context

  • Target ICP
  • Personas
  • Competitive intel
  • Differentiators
  • Win/loss reasons
  • Messaging
  • Customer quotes

01

Claude misses important market signals

Rival social posts, new ads, fresh customer reviews. Claude can't read them.

Calven reads gated sources to show where rivals spend and what their customers say.

Competitor dossier: Northwind

Written by Claude from web search

Overview

Northwind sells month-end close software to mid-market finance teams. In October it added multi-entity consolidation, capped at five entities per plan.

Recent moves

DateMove
Nov 4Native SAP S/4HANA integration
Oct 14Consolidation for up to five entities
No ad data found

What customers say

  • Rated 4.4 on G2 from 212 reviews.
  • No dated reviews found.
  • Pricing starts at about $1,500 a month.

Sources

  • northwind.com/blog/introducing-sapPage not found
  • techcrunch.com/2023/05/northwind-series-bFrom 2023
  • linkedin.com/company/northwindCouldn't read
  • g2.com/products/northwind/reviewsNo dated reviews

02

Claude guesses. With Calven, every fact has a source.

Claude cites the web pages it reads. Calls and deals it cites only if your skill pulls them in. The price a buyer was quoted is in neither, so it fills the gap with a guess.

Calven ties each fact to the page, call or deal it came from, with the date and the words. You can check every line in one click.

03

Claude rewrites the whole document on every run

Claude alone

A scheduled task in Claude writes a new document on every run, and the changes get lost in the rewrite.

Battle card: NorthwindEdited by scheduled task

Northwind: Competitive Battle Card

Refreshed November 10

Their strengths

Northwind offers a bigger partner network, a native NetSuite app and an SAP integration.

Who they are

A month-end close platform aimed at mid-market finance organizations.

Our advantages

Coverage. Reconciliation spans all entities at once.

Speed. Customers go live within days.

Trust. Each number is linked to its source.

Questions to plant

"How many entities did your last customer close?"

"How long did setup take them?"

Every linechanged. One new fact. Find it.

With Calven

Calven's agent adds the one thing that changed and sends its owner a draft. Your skills read the approved card.

Battle card: NorthwindDraft · 1 change

Battle card: Northwind

Last updated November 3

Overview

Northwind sells month-end close software to mid-market finance teams.

Where we win

  • Reconciliation across every entity, not one ledger at a time.
  • Live in days, not months.
  • Every number traces to its source.

Where they win

  • Larger partner network.
  • Native NetSuite app.
  • Native SAP integration, launched November 4.Source: northwind.com/changelog · detected November 5

Landmines to set

  • Ask how many entities their last customer closed.
  • Ask how long setup took them.
1 lineadded. 14 untouched.
The same card after the same competitor launch. Illustrative example.

04

Claude reads the notes. Calven interviews the buyer.

Claude analyzes the win/loss notes and transcripts you give it. Most deals never get any.

With Calven, the Win/Loss Agent picks up every deal that closes in your CRM. It finds the buyer, sends the invite and runs the survey or AI voice interview. Nobody schedules a call or chases a reply.

Halden Group · Opportunity

Closed lost
  1. Deal closed in SalesforceNov 3

    Closed lost to Northwind

    Win/Loss Agent started the loss surveyAuto

  2. Buyer found and invitedNov 3

    Erin Walsh · Head of Finance

    Invite sent
  3. AI voice interviewNov 6 · 14 min

    • Product4.6 out of 5
    • Sales team4.2 out of 5
    • Pricing & legal1.8 out of 5
    “We preferred your product. Our lawyers were still marking up your contract when Northwind’s was signed.”
  4. Loss report and patternNov 6

    Third loss this quarter that stalled in contract review, after the buyer had picked you.

    Read the full loss report

05

The latest facts in every AI tool. No prompt needed.

Calven's agents keep the ICP, positioning, messaging and competitive intel current. Calven MCP serves the approved version to everyone in the company.

Nobody has to type "check Calven". Claude knows what Calven MCP holds and pulls it in whenever a question needs it.

  • Online research

    Competitors, news, market

  • Documents

    Decks, reports, briefs

  • Call notes

    Meetings, transcripts

  • CRM

    Deals, contacts

  • Win/loss

    Surveys, interviews

9 agents extract the insights

Calven MCP

Your GTM source of truth

  • Competitor intel
  • Win/loss surveys
  • Customer voice
  • Market trends
  • Market opportunities
  • ICP profile
  • Positioning framework
  • Persona definitions
  • Product brief
  • Messaging framework
  • Pipeline insights
  • And many more…

Claude serves every team

  • Product marketingFor the launch brief: every customer call that asked for multi‑entity close, and why.
  • SalesWhat won our finance deals against Northwind? Prep my Helio Bank call around it.
  • MarketingWrite LinkedIn ads in our customers' words and rank them as a Head of Finance.
  • ProductRank the product gaps that cost us deals last quarter, with the buyer quotes.
  • Jordan Lee10:14 AM

    Does Northwind support multi-entity close yet? Prospect is asking.
  • Priya Shah10:31 AM

    Just got hold of Northwind's new pricing deck. Sharing for reference.
    PDFNorthwind pricing 2027.pdf
  • Sam Ortiz10:52 AM

    Do we have any intel on Quillpoint? They just showed up in a deal.

Jordan Lee10:14 AM

Does Northwind support multi-entity close yet? Prospect is asking.

CalvenAgent10:14 AM

Partly. Northwind added consolidation in October, capped at 5 entities per plan. You have no cap.

Sources: Northwind dossier · Battle card: Northwind

06

Give Calven a channel.
It answers every question in it.

In Slack or Microsoft Teams, it answers in the thread with sources, starts tasks and tags a person when it doesn't know. It also takes direct messages and posts a weekly digest.

07

A skill runs when asked. Agents are autonomous.

Leadership changes the positioning, and every battle card is out of date. With Claude alone, they get updated only if someone built a skill and a trigger for that case in advance.

In Calven, the agents know what is built on what. Publish the positioning and they draft every update themselves.

Claude alone

Customize › Skills
  • Competitor watchScheduled: Mondays, 4 AM
  • Battle card builderRuns on: new competitive signal
  • Messaging frameworkWaiting to be asked
  • ICP builderWaiting to be asked
Refresh battle cards when positioning changesNot built, no trigger. Nothing happens.

With Calven

Positioning Framework updated

  1. Messaging Agent

    Update the messaging framework

    Draft ready: 3 sections changed
  2. Competitive Intelligence Agent

    Update all 14 battle cards

    14 battle cards updated
  3. ICP Agent

    Check the impact on the ICP

    No change needed
Posted to #product-marketing: 15 updates are waiting for your review.

Spend your hours on the work only you can do

Writing your first skill is quick. Keeping it and everything around it working is a job that never ends. Connect Calven and that job is ours.

What it takes to keep home-built Claude skills workingFull guide →

  • Up to

    28 h

    of upkeep a month

  • 4 to 8 h a month

    Quality drift

    Output slides as the market and the product move

  • 4 to 6 h a month

    Model updates

    Every skill is re-tested when the model changes

  • 2 to 4 h a month

    Broken sources

    Sites redesign, links break, APIs change

You built what was in my head. I'd rather pay someone who's already built it than keep building and maintaining it myself.

Director of Product MarketingB2B software company

Nine agents that work as one team

Each agent runs on product marketing best practices, and all nine share one picture of your market. What one agent learns, the others build on.

Written and tested

115skills

Product marketing best practices in every one. Hover to see them.

Research & Intelligence61

Target Audience23

Positioning & Messaging31

Ready from day one

Everything works out of the box. Nothing to build, configure or prompt before the agents start.

Tested and improved

Every skill is tested before it ships, and improvements reach you automatically. Nothing to maintain.

Connected

Change the ICP and the agents update personas, positioning and messaging to match.

  • Salesforce
  • HubSpot
  • Gong
  • Zoom
  • Slack
  • Google Workspace
  • Notion
  • Confluence
  • Apollo.io
  • ZoomInfo
  • Asana
  • Claude
  • ChatGPT
  • Gemini
  • MCP

Plugged into your stack

Your CRM, call recordings, Slack and Google Workspace, plus every AI tool over MCP.

The details, row by row

What Claude does on its own, and what Calven adds. Claude's side was checked in November 2026.

Market and buyer evidence

Competitor websites, docs and news

Manual

Claude alone: Reads the pages when a run starts. Comparing with last time is yours to build

Built in

Claude with Calven: Crawled weekly or daily and compared with the last copy

Competitor LinkedIn posts

Not available

Claude alone: LinkedIn blocks Claude from opening the page. Search shows a few old snippets at best

Built in

Claude with Calven: Each competitor's new posts, checked every 3 days

Competitor ads on LinkedIn and Google

Not available

Claude alone: Web fetch can't load either ad library

Built in

Claude with Calven: New ad copy and creative, every week. Text in image ads is read too

Reviews on G2, TrustRadius, PeerSpot and Gartner Peer Insights

Not available

Claude alone: Search shows a score and a review count, not the reviews. G2's connector needs a paid G2 plan

Built in

Claude with Calven: Ratings every month, with the month's new reviews summarized

Competitor hiring

Manual

Claude alone: Reads the postings on each run. Comparing month to month is yours to build

Built in

Claude with Calven: Compared month to month: hiring surges, new countries, leadership roles

AI share of voice

Not available

Claude alone: No built-in way to ask other AI assistants. It takes a paid data service and your own schedule

Built in

Claude with Calven: Six assistants asked how they describe you and who they name, every month

Sales calls

Manual

Claude alone: The calls someone asks about, through their own connector

Built in

Claude with Calven: Every call from Gong, Fathom, Fireflies and Granola

CRM deals

Manual

Claude alone: The deals someone asks about, through their own connector

Built in

Claude with Calven: Every deal in your CRM, synced daily

Buyer interviews

Not available

Claude alone: Analyzes notes and transcripts. It doesn't invite or interview buyers

Built in

Claude with Calven: Invites the buyer after a deal closes and runs the survey or AI voice interview

Documents you can trust

A source behind each fact

Limited

Claude alone: Web answers cite the page inline. Calls, deals and finished documents depend on your skill

Built in

Claude with Calven: Each fact carries a numbered citation to the page, call or deal

Updating one section

Manual

Claude alone: A scheduled run writes a new document. A section edit needs someone to ask

Built in

Claude with Calven: Changes the affected section and shows the edit

Your hand edits

Limited

Claude alone: Kept only if the skill starts from your current file. No conflict check

Built in

Claude with Calven: Kept. A conflicting agent edit is flagged for review

Review and publishing

Manual

Claude alone: No draft or published state. Whoever ran it decides

Built in

Claude with Calven: Drafts go to the document's owner. You set the autonomy

Where your team gets the answers

One approved version

Manual

Claude alone: A shared Project or Drive file that someone keeps current

Built in

Claude with Calven: One published version for the whole company

Other AI tools

Manual

Claude alone: Skills are portable. The facts behind them are copied into each tool by hand

Built in

Claude with Calven: The same approved facts over MCP in Claude, ChatGPT, Cursor and GitHub Copilot

Slack

Limited

Claude alone: Answers in channels on Team and Enterprise plans, billed per use

Built in

Claude with Calven: Owns a channel: answers with sources, starts tasks and tags a person when it doesn't know

Microsoft Teams

Not available

Claude alone: Claude's Teams app is still on a waitlist

Built in

Claude with Calven: The same channel agent as in Slack, with direct messages and a weekly digest

Reacting to change

What starts the work

Manual

Claude alone: A person, a schedule or a Slack channel watch that someone set up

Built in

Claude with Calven: New evidence: a changed page, a new call, a closed deal, a published document

One change, many documents

Manual

Claude alone: Only the chains someone built in advance

Built in

Claude with Calven: Publishing one document queues the updates that depend on it

Running it

Getting started

Manual

Claude alone: Every source, schedule and check is yours to build

Built in

Claude with Calven: Agents working on day one. You write the skills that are yours

Upkeep

Manual

Claude alone: 15 to 28 PMM hours a month, by Calven's estimate

Built in

Claude with Calven: The plumbing is Calven's job. Your skills stay small

What a full build takes to keep running

Quality drift4 to 8 hours per month
Model updates4 to 6 hours per month
Broken sources2 to 4 hours per month
Sales feedback2 to 4 hours per month
Security patches2 to 4 hours per month
Usage costs1 to 2 hours per month
Total15 to 28 hours per month

Claude with Calven, answered

Can we keep our Claude skills if we use Calven?

Yes, and keep building them. Connect Calven MCP and every skill you wrote reads the approved, current ICP, positioning, messaging and competitive intel, with sources, instead of whatever someone pasted in. Claude is where your team asks and creates. Calven is where the facts come from.

We already use Claude. Why add Calven?

Claude gives your team a place to ask and create. Calven adds the seven things on this page: the sources closed to web search, a source behind every fact, one-line updates with approval, buyer interviews, one approved version in every AI tool, an agent that owns a Slack or Teams channel and agents that react to change. You don't build or maintain any of it. Calven Professional costs $9,600 a year for the whole team.

Which Claude skills should we build ourselves?

The ones that carry your format and your judgment: the pitch deck in your brand template, the copy review against your style, the call prep your sales team likes. Each is a prompt, a template and Calven MCP for the facts. Leave the research, the monitoring and the approved documents to Calven.

Can I rebuild Calven with Claude skills?

Part of it. Skills draft battle cards, summarize a competitor and cluster win/loss notes well. They don't interview your buyers, read the sites that are closed to web search or keep one approved version current for the whole company. Those jobs take software running around the model. Building and running that yourself takes 15 to 28 PMM hours a month by Calven's estimate, ~$15,000 to $29,000 a year at about $86 an hour.

Does the same apply to ChatGPT?

Mostly. ChatGPT has skills, projects and scheduled tasks too, and Calven MCP connects to both. ChatGPT's search reads more of LinkedIn and the review sites than Claude's, but not the ad libraries, and everything else on this page applies the same way. This page uses Claude in its examples.

Can't Claude read LinkedIn through a browser extension?

Claude in Chrome can open pages in a browser where you are signed in. LinkedIn's User Agreement prohibits automating activity on the site, and Anthropic leaves compliance with each site's terms to the user. It also runs on one person's account, and only when that person asks.

What about scheduled tasks in Claude Cowork?

A scheduled task runs a saved prompt hourly, daily or weekly. Each run is a new session, so comparing with last week works only if you build the task to save and re-read its own notes. Each schedule belongs to one person's account. Five PMMs with five schedules produce five battle cards for the same competitor, none of them approved.

How fast does Calven start working?

On day one. Add your website and your competitors, and the agents build the first ICP, positioning and competitor dossiers on their own. The 14-day trial needs no credit card and no sales call.

Two weeks trial. You'll only need a day.

A day is all it takes to see what the agents can do for you.
You won't go back.

Start free trialPrefer a walkthrough first? Book a demo