MCP Tools Documentation

Call this FIRST. Returns your journey progress, active goals, coaching streak, momentum profile (planner vs executor), integrations, wins, and what to do next. The entry point for every session.

Describe what you need in plain language and it routes to the right tool. "How do I get customers?" routes to suggest_next_step + get_coaching. "Is my positioning good?" routes to mark_score_positioning.

Returns the single most important action to take right now based on your journey progress. Includes a real startup example for motivation.

Checks if you are a planner or executor. Scores planning (positioning, personas, offers) vs execution (interviews, outreach, logged actions). Returns a motivational nudge and a 5-10 minute quick-win action.

Create a free account directly from your terminal — no browser needed. Returns an API key instantly.

Guided setup for a founder who has just arrived, or for a new product. Walks the journey one plain-language question at a time and assumes no marketing vocabulary at all. Give it your website, landing page or app listing and most of the answers are read back off your own page — the name, what it claims, your colours, fonts and logo, and where else you already appear — so you confirm rather than compose. Nothing read this way is saved until you say yes, and every proposal shows the place it came from. A step that needs a real conversation, like talking to customers, can be put aside and the flow carries on.

Where you are in setup and the single next question to ask. Returns what is done, what was put aside, and anything read from your site still waiting to be confirmed. Progress is derived from your actual records rather than a stored cursor, so starting in the browser and finishing in your terminal picks up exactly where you left off.

Record that a setup step is answered and get the next question. Call it after the answer has been saved with the tool that step names — it advances the flow rather than writing the record, and it will tell you if nothing was actually saved.

Put a setup step aside and move on. Most often customer interviews, which need a real conversation and cannot be done at a keyboard. This is a normal outcome on day one, not a failure: the step comes back at the end as something waiting on the real world, with what it needs before it can be answered.

Check your API usage — calls today, this month, quota remaining, and which tools you use most.

Define your product — the problem it solves, the solution, differentiation, target keywords for SEO, and presence channels. This is Step 1 and everything else builds on it.

Edit product details — name, description, problem statement, target keywords, presence channels. Partial update — only send fields you want to change.

Show all your products with their IDs. Use this to get a productId for other tools.

Log a customer interview — who you talked to, their pain points, current solutions, budget, and whether they would try your product. Aim for 5-10 before positioning.

Add more details to a previously logged interview.

Show all customer interviews, optionally filtered by product.

Add an action item from research feedback — e.g., "create a community group" or "follow up on pricing concern".

Show open research action items.

Mark a research task as done or update its title.

Remove a research task.

Write your positioning statement — target customer, core value, differentiation, and elevator pitch. Best after 5+ interviews.

Refine your positioning statement.

Show all positioning statements.

Define your ideal buyer — job title, pain points, channels they use, budget, objections, fears, and aspirations.

Edit a buyer persona.

Show all buyer personas.

Package your product into a specific offer — pricing, guarantee, bonuses, payment terms, and positioning statement.

Change price, guarantee, bonuses, or description.

Show all offers with pricing.

Plan your go-to-market — which 2-3 channels to use, weekly targets, and 90-day goals.

Change channels, targets, or monthly goals.

Show GTM plans and channels.

Map your sales funnel — stages from awareness to paid, with conversion targets.

Update funnel stages, targets, or current metrics.

Show all funnels with stages and metrics.

Track daily marketing work — emails sent, calls made, posts published, demos given. Builds your execution history for coaching.

Edit an existing activity log entry.

Delete an activity log entry.

Log marketing work in plain language — "I sent 10 cold emails today", "I posted on Reddit", "I had a demo call". It works out the activity type and quantity for you, so logging a day costs a sentence instead of a form.

See what you have filled in and what is missing across all 8 steps, with completion percentages.

Personalized weekly coaching plan — diagnoses bottlenecks (outreach>reply, reply>meeting, meeting>close), gives 5-7 actions with industry benchmarks. The core coaching engine.

Your weekly coaching plan with specific tasks, their priority and status, goal progress, and industry benchmarks.

Mark a coaching task as done. Record the outcome: worked, partially, did not work, or skipped. Builds your learning history.

Your business stage (Idea, Validating, Positioned, Market-Ready, Launching, Growing) with a readiness score (0-100) and priority actions.

Score how market-ready one product is across 5 dimensions: Definition, Validation, Market Fit, GTM, and Traction.

Compare market-readiness across all your products.

Smart reminders about stale data, missing fields, momentum loss, and milestone celebrations.

Everything about one product in a single call — definition, research, positioning, persona, offer, GTM, funnel, execution, maturity, and nudges.

A full marketing health check. Reads all 8 steps, then names the bottlenecks, the places where two steps contradict each other, and the specific next actions. This is the answer to "how am I doing?" and "what should I focus on?".

A 0-100 score for each step — product, research, positioning, persona, offer, GTM, funnel, execution — plus overall readiness. The quick version of a diagnosis.

Advice for one step, cross-referenced against every other step. Not "write a better persona" but which field is empty and which interview already says what belongs in it.

Suggestions for a step taken from your own data — pain points from research, budgets from personas, competitor pricing, benchmark conversion rates. Call it BEFORE writing positioning, an offer, a GTM plan or a funnel, so the draft matches what you actually found instead of guessing.

Your whole strategy as one portable document — product, research, positioning, personas, offer, GTM, funnel and execution. Ask for the CLAUDE.md format to drop it straight into a coding agent’s project context.

Complete marketing context — all data in JSON, Markdown, or summary format. Use to export your marketing state to other tools.

Record a competitor with their presence, offerings, pricing, strengths, weaknesses, and where you can beat them.

Update competitor data — new pricing, features, or market moves.

Show all competitor analyses with threat levels.

Structured comparison of your product vs competitors — positioning, pricing, presence, and opportunities.

Agency-grade website audit — tracking, SEO, CTAs, trust signals, lead capture, messaging alignment, and offer presentation. What agencies charge $2,000-$5,000 for.

Agency-grade competitive audit — SWOT analysis, feature parity, pricing position, channel gaps, vulnerabilities, and moat evaluation.

Questions for a discovery call, written for your product — rapport openers, pain discovery, budget questions, and an explicit DO NOT ASK list. The list is the useful half: most first interviews are spoiled by pitching.

Save an interview as structured data — who you spoke to, the pain in their own words, their budget, and whether they would try it. This is what positioning, personas and synthesis all read from.

Find the pattern across every interview — pain points ranked by how often they came up, budget signals, the strongest positioning angles, and whether you have had enough conversations to be confident yet.

Agency-grade journey audit — data quality, cross-step consistency, execution velocity, risk assessment, and conversion readiness.

Reads the raw HTML of a site, not just the words on it — meta and Open Graph tags, semantic structure, heading hierarchy, image alt coverage, render-blocking scripts, JSON-LD, and outbound links missing rel="noopener". Catches what looking at a page cannot.

Set a measurable goal linked to a product — "100 customers by October", "$1k MRR", "1000 website users". Supports auto-tracking via integrations.

Show all goals with progress percentages.

Log progress toward a goal — update the current value.

Create a UTM tracking link or promo code to measure which channels bring customers.

Show all tracking links and promo codes with click/conversion stats.

Add a potential customer to your pipeline. Track them from cold outreach through conversion.

Update status, temperature (cold/warm/hot), tags, or details. Track replies, meetings, and conversions.

Show leads with temperature, tags, and follow-up alerts. Filter by status or temperature.

Remove a lead from your pipeline.

Add tags to leads for segmentation — e.g., "Q3 campaign", "Stripe customers", "hot prospects".

Import leads from connected platforms — Stripe customers, HubSpot contacts, Mailchimp subscribers, Gumroad buyers, Slack members.

Schedule a meeting with a lead — creates a calendar event and updates the lead with the meeting link.

Enrich a company or contact with data from Clearbit, Apollo, or ZoomInfo.

Create or update contacts in HubSpot or Salesforce.

Search leads, deals, contacts, or pipeline data in your CRM.

Add to your content calendar — blog posts, tweets, emails, LinkedIn posts. Track title, channel, date, and status.

Create or list content calendar items in one tool.

Show planned and published content.

Change title, status, channel, or scheduled date.

Remove a content calendar item.

Auto-fill a marketing template (cold email, LinkedIn message, landing page) with your real product and persona data.

Set up an A/B test — hypothesis, variants, target metric. Structured experimentation.

Record what happened — which variant won, learnings, and whether to scale or pivot.

Show all experiments and their status.

Change hypothesis, status, or add learnings.

Add or list testimonials, reviews, and social proof.

Show all testimonials.

Edit a testimonial.

Remove a testimonial.

List all blog posts from The Mark Platform blog. Filter by tag.

Get the full content of a blog post by slug — includes author details and social links.

Everything needed to write a campaign in one call — persona, positioning, offer, channel best practice, copy formulas, and the DO/DON’T list for that channel. Covers cold email, LinkedIn, blog, community posts and Twitter threads.

Channel-specific outreach built from your persona, positioning and offer — templates, the PAS/BAB/FAB formula that suits the channel, and the rules that stop it reading like a pitch.

Turn a sequence into a CSV your sending tool can import — Saleshandy, Instantly, Lemlist, or plain CSV. Subject, body and follow-ups, formatted per tool.

Send through your own connected email platform (Resend, SendGrid, Lemlist, Mailchimp). Auto-logs the activity, and can record it as an interview request. Nothing is ever sent from an address you do not own.

Plan a marketing campaign — paid ads, cold outreach, or organic. Uses templates like "first-10-customers" and "product-hunt-launch". Auto-generates ad copy from your data.

Campaign performance — spend, revenue, clicks, conversions, ROAS, CPA, and alerts.

Validate budget and performance limits — flags overspending or underperformance.

Generate ad copy or marketing text using PAS, BAB, or FAB formulas. Pulls your product, persona, and brand context automatically.

Store an ad copy variant for a specific platform (Meta, Google, LinkedIn). Used by AI agents and campaign launching.

Show all stored ad copies with platform, format, and status.

Define brand guidelines for AI-generated ad content — tone, visual direction, colors, and rules.

Show all brand style guidelines.

Full ad context — all copies, styles, and live campaign snapshots in one call.

Set up a Meta pixel on the connected ad account so conversions can be tracked. Reuses an existing pixel rather than creating a second that would split your data.

The exact pixel code to install on your site, with your pixel ID and one event block per funnel stage already filled in. You paste it into your own site — we cannot install it for you.

Report a conversion to Meta so ad reporting shows what your spend produced. SHARES YOUR CUSTOMERS DATA WITH META: email, phone and name are SHA-256 hashed before sending, but hashing is designed to let Meta identify the person. Off by default; needs the Privacy Centre consent AND the integration permission, separately.

Show exactly what an ad campaign would look like before creating it — targeting, ad copy, budget and estimated spend. Creates nothing and touches no ad platform. Returns the previewHash that create_ad_campaign requires.

Create a real campaign, its targeting and its first ad on Meta Ads or Google Ads. Everything is created PAUSED and spends nothing until you resume it. Requires the previewHash from preview_ad_campaign and the separate "Create Campaigns" write permission.

Show running ad campaigns from Meta Ads, Google Ads, or LinkedIn Ads.

Pause a live campaign so it stops spending. Takes effect on the ad platform immediately. Needs the per-platform write permission enabled on your integration.

Resume a paused campaign. Spend restarts as soon as the ad platform picks it up.

Change a live campaign's daily budget, in whole currency units of the ad account.

Show agency invitations waiting for your answer — who invited you, which agency, what role, the projects you would get access to, and how many days are left before the invite expires.

Join the agency and get access to its projects. Only usable by the person the invite was addressed to — a forwarded link does not let someone else claim the seat.

Refuse an invitation. Marks it declined so the agency owner sees it was answered, rather than watching it expire silently. Nothing is shared with the agency.

Your brand identity for a product — archetype, voice attributes, contrast pairs, colours, typefaces, origin story and completeness score.

Define how your product sounds and looks. Start with an archetype and three contrast pairs ("direct, but never blunt") — those two fields alone change every draft.

Change any brand field. Partial — only what you send is touched.

Remove a product's brand identity.

The prompt-ready voice block — everything the brand knows, written as instructions. Paste it into any AI tool to get on-brand output there too.

Score a draft against your brand rules. Checks what is mechanically checkable — banned words, banned claims, emoji policy, contractions, filler — not whether it feels right.

The 12 archetypes with what each implies for voice, plus seed traits and contrast pairs so you edit rather than invent.

Your on-brand / off-brand copy examples, and which surfaces still have none. These teach a generator how you actually write — they lift generated quality more than any other brand field.

Save a copy pair: the same message written the way you write it, and the way you do not. ADDS to the list, never replaces it. The contrast is where the rule lives.

Edit a saved copy sample, or retire it.

Retire a copy sample so it stops shaping output. Soft by default.

Where your brand needs to become something physical or fixed — business cards, stickers, packaging, signage, merch, slide templates, email signatures, invoices, app store listings, booths, vehicle livery. Ordered cheapest first, filtered to what applies to your kind of product, and honest about which ones are not worth doing yet.

The production brief for one artefact — real dimensions, bleed, colour space and material notes, every design decision traced to the brand or persona field that made it, and a ready-to-paste design prompt. We write the spec; we do not design it or hold the files.

Words to use, words to avoid, claims never to make — each with a severity and a reason.

Add one rule without rewriting the list. Hard rules are absolute; soft rules are preferences.

Change a rule's text, severity or rationale.

Retire a brand rule.

Live numbers from a connected game store, graded against real indie benchmarks. Steam gives review score, review count, live players and price with an App ID alone — no key needed — and adds wishlist adds, deletes, purchases and revenue once a Financial API key is connected. itch.io gives page views, downloads, purchases and earnings per game. It also reports what CANNOT be connected and why: Epic publishes no self-serve API and the console platform holders gate everything behind NDA portals, so those numbers can only be read by hand. A founder who cannot find the Epic integration should learn it does not exist rather than assume they missed it.

Builds the streamer outreach list, which is otherwise an hour of manual searching every time you need one. Searches Twitch for people currently streaming a comparable game and returns the channel, live viewer count, language and stream title. Filters out channels too large to read their own email by default, because the useful target is mid-sized — handing a founder the ten biggest channels on Twitch is handing them ten people who will never reply. Search a game LIKE yours, not your own: nobody is streaming an unreleased game.

Where to publish a game, and in what order. The portals, stores and embedded surfaces that will actually host it — Poki, CrazyGames, itch.io, Newgrounds, Coolmath Games, the aggregators, YouTube Playables, Discord Activities, Telegram Mini Apps, Google Play and the App Store — with what each pays, what it requires, what it costs and who is there. The thing most developer-founders have never been told: a mobile-shaped casual game can be published on the web for nothing and earn an ad revenue share, with no $99 Apple membership, no twelve testers for fourteen days and no install. The field that decides the strategy is EXCLUSIVITY rather than the revenue split — Poki pays best and requires web exclusivity, while an aggregator syndicates to thousands of sites and therefore rules Poki out permanently. That decision is irreversible and is usually made first, by accident. Commercial terms carry their provenance: what a platform publishes itself is separated from unconfirmed third-party figures, because choosing a distribution partner on a made-up percentage is a real decision on a false input.

Where a game finds its players. Returns the discovery surfaces that actually sell games — store page, short-form gameplay video, streamers and creators, genre subreddits, storefront festivals, your own Discord, game jams, press, app store search — with what each one is GOOD AT, what it costs in real hours, and what a realistic outcome looks like. That last part is the point: the most common way a game founder wastes six months is judging a reach channel on conversions, or a retention channel on reach, and abandoning the one that was working.

The launch timeline as dated beats: announce and open the store page, build the wishlist through the long middle, ship a demo, enter a festival, the six weeks before launch, launch week, and the tail. Each beat carries EXIT CRITERIA — what has to be TRUE before the next one is worth starting — which is the part most launch checklists omit. A founder who announces on schedule with 300 wishlists and moves on on schedule is executing a plan that has already failed, and nothing else tells them. A launch date is optional; without one the beats come back sequenced rather than dated.

What creatives a game needs, at real dimensions, and honestly which half we can make. Capsule, screenshots, announce and launch trailers, GIF loops, vertical shorts, devlog thumbnails, milestone cards, press kit, app icon — each with what its first second has to accomplish and the mistake specifically made on that asset. A game's proof is FOOTAGE, so most of these are capture specifications: we set the capsule and the announcement card, and the four seconds that make somebody stop scrolling is something only you have.

The benchmarks for a game, which are not the platform's usual ones: wishlists at launch, wishlist-to-purchase in week one, store-page visit to wishlist, review score, refund rate, demo-to-wishlist, and what share of year-one revenue arrives in week one. Metrics that come back null are ones your platform genuinely does not have — itch has no wishlist — and they are named as absences rather than returned as zero.

Why the number is bad and what to change, in order, with what NOT to do yet. Covers the six failures a game actually has: nobody reaches the page, people reach it and do not wishlist, the demo produces no wishlists, refunds are high, the review score is dragging the listing down, and sales stopped a week after launch. The diagnosis is almost always further up the funnel than the symptom — which is why every rule carries a "do not do yet", and why the first one is usually "do not buy traffic for a page that does not convert".

The weekly rhythm, with what each slot is FOR. A CLIP is for strangers and is about the game; a DEVLOG is for people who already care and is about the work. Conflating them is why most build-in-public accounts post engineering progress to an audience of nobody. Returns the realistic weekly time cost and the genre communities to post in, so you can say no to it honestly rather than agreeing and quietly stopping in three weeks.

Step-by-step guides to the twelve channels that sell games: the Steam store page, Steam Next Fest, Discord community, gameplay shorts, streamer and creator outreach, press and key distribution, genre subreddits, devlogs, itch.io and game jams, crowdfunding, mobile store optimisation, and web game portals. Each has numbered steps with real time costs, what success looks like, and when to pivot away. A separate pool from the general playbooks on purpose — a game founder should never be pointed at LinkedIn cold outreach.

The first step in making any creative, and the one the platform used to skip entirely. Builds the angle map for your product: the 8-15 genuinely different arguments your ads could make, each sourced from something a real customer actually said in an interview, a testimonial, or a recorded objection. An ANGLE is what an ad argues and to whom; hooks, formats and art direction are all downstream of it, which is why two ads with the same angle and different layouts are one test rather than two. Angles with no evidence behind them are dropped rather than padded out, because an invented angle looks decided, renders at full size, and becomes a claim you now own. Ask for a test set and it picks genuinely different angles across territories instead of variations on one.

Turns one angle into a brief you write against — the step where copy actually gets written. It hands over the the phrasing customers actually used, what the opening has to accomplish, the character ceiling and job for every line, and the framings that would specifically kill that angle. It deliberately does not write the copy for you: your AI tool is the writer, and the quotes are raw material rather than lines to paste. If you generate a creative without written copy it falls back to database fields and the response says so — that fallback is what made every creative from this platform read the same.

Commission a video or image creative. It does NOT generate one for you — it hands you a brief and an empty canvas, and your AI tool makes the creative. This replaced a catalogue of seventeen templates whose output was measurably always the same thing: type on a coloured ground, in one of fifteen pre-drawn boxes, so a founder generating ten creatives over a quarter got one creative with ten sets of words in it. You receive the argument (an angle sourced from what your customers actually said), your brand as MATERIAL rather than as a cage, the real mechanics of the surface you are posting to, a fresh provocation drawn from a 48-card deck, six non-negotiables, nine explicitly stated freedoms, and a Remotion project whose main component is deliberately empty. Your tool then writes that component: any composition, any layout, any motion, any palette — including colours outside your brand palette when the idea earns it — and as much or as little copy as the idea needs, including almost none. There is no template to pick and no slot to fill. Two commissions for the same product should not look alike. Only six things are fixed, and four of them are legal or accessibility obligations: copy stays out of the regions the platform draws its own UI over, text clears the contrast and size floor, every number traces to something real, every image ships its own alt text, every render carries a C2PA manifest, and the Remotion licence position is stated — you become the licensee, free up to 3 people. Brand rules travel with their rationales and are reported rather than enforced: review_creative will tell you when something reads off-brand, and it will not refuse to ship over it. Generated projects go in a `.mark/` scratch folder by default — the whole thing can be regenerated from the seed, so it is a cache rather than something to commit — and your assistant asks once whether to gitignore it. Pass `outputDir` to keep it somewhere permanent instead.

The deck a commission draws from: 48 creative stimuli, in four kinds. `steal` imports a visual grammar from a discipline that is not advertising — transit signage, a scientific diagram, a receipt, flat-pack instructions, a protest placard. `constraint` removes or forces something: seven words in the whole creative, no type until the last beat, only the customer may speak, none of your brand colours, ugly on purpose. `tension` forces two things together that do not belong: lead with the objection, the mundane shot like a trailer, honest to a fault. `era` dates it precisely: Swiss 1960, punk zine 1977, blueprint 1910, risograph 1986, arcade 1983. None of them can be applied mechanically, and that is the point — interpreting one is work only your AI tool can do, and two interpretations of the same stimulus should not look alike. Each carries what it unlocks and how it fails, so you can decline one that would wreck a brief. The deck is shuffled per product and walked rather than rolled, so it does not repeat for 48 commissions.

The shape of the canvas before you commission anything: every output spec at real pixel dimensions, the safe areas each platform draws its own interface over (a Story and a Reel are the same pixels and reserve different regions — getting that wrong hides your call to action rather than merely crowding it), the supported lengths from a 6-second bumper to a 60-second explainer, the six non-negotiables, and the nine freedoms that are explicitly yours. Read it when planning a campaign across several surfaces.

Review a creative before you publish it, and get a verdict: publish, revise, or throw it away and start again. It grades against four things at once — your brand rules, the platform (hook against the swipe window, length against the sweet spot), what the creative is FOR (an awareness piece is marked down for a hard sell; a performance ad is marked down for having no ask), and the content itself (copy cut off mid-line, numbers nobody can source, a rendered file with no Content Credentials on it). It reads CREATIVE.json, the manifest your tool fills in after writing the component, so an empty manifest returns cannot_review rather than inventing findings about nothing. Brand rules are reported and never block: only the six non-negotiables do, and four of them are legal or accessibility obligations. Every finding says why it matters and what to do.

Which open-source simulation tool answers your question, and whether you need one at all. Covers CalculiX, Elmer, Code_Aster, OpenFOAM, SU2, FEniCSx, Gmsh, FreeCAD FEM and ParaView, with what each one solves, what it does NOT solve, its licence, the honest runtime, and how its output becomes frames in a Remotion project. The platform runs none of it and ships no part of these projects — you choose what to install and run it on your own machine. The distinction that decides the answer: OpenFOAM is CFD, and a stress contour on a loaded part is FEA. It leads with the answer that is usually right, which is that you probably do not need a solver: most technical creatives fail on a missing support or an unlabelled stress measure, not on a missing simulation.

What makes a force diagram, a stress contour, a mode shape or a flow field correct — the checks an engineer in your audience runs in the first second. Call it before drawing one, and pass the declaration back to review_creative so the mechanical checks run: a load with no support, a peak stress quoted at a sharp corner where stress is theoretically infinite, an unnamed stress measure, a number with no material behind it, element counts that cannot describe the mesh drawn. An illustrative figure is honest work and says so in the frame; the same figure presented as a solver result is a claim you did not make and now own.

Check whether a rendered image or video actually carries C2PA Content Credentials, and show what they declare — which agent or model made it, and whether the file says AI was involved. Run it after every attach step and before publishing. Attaching credentials and having credentials that read back are different facts: a signing step that fails halfway leaves a file that looks finished. Works on any file, not only ones this platform generated. Needs the MCP server running locally, because the file is on your machine and never on ours.

The three things a creative can be for — building awareness, getting a tracked click, or both — and what each is graded on. Worth asking before commissioning: the same clip is a success as an awareness piece and a failure as a performance ad, because a hard call to action is what makes one work and what makes the other read as an ad and get scrolled past.

Find creatives to LOOK AT before making one, then turn what you saw into something commissionable. Research used to be text only — search result titles, URLs and a written summary — so briefs about creative work were being written by something that had never seen a creative, and it showed. Call it once and it returns search angles built to find creative work rather than products, real pins with an image URL on each, and the questions to answer while looking. Open the images, call it back with what you saw, and it returns concrete offers: a shape, a length, a provocation and the exact commission call for each — every one traced to a reference that was actually opened.

How a specific surface actually behaves — video or image. The hook window before the median viewer swipes, what the ranking system rewards, whether sound is on by default, caption and alt-text practice, posting cadence, which funnel stage it serves, and the mistakes that cost the most there. Also says when NOT to use a surface, which is more useful than another tactic. Read it before commissioning for a platform you have not worked on: a creative that is wrong for the surface has failed before production quality is relevant.

Show all connected integrations and their sync status. Always call this before assuming a platform is unavailable.

Pull fresh data from a connected platform. Works with every connected integration.

Get real numbers — Stripe revenue, GA4 visitors, GitHub stars, PostHog events. Time-series data from any connected platform.

Which marketing channels are working? Aggregates data by GTM channel — sessions, conversions, revenue.

Cross-platform conversion totals rolled up into awareness > interest > consideration > paid. For per-stage drop-off on your own funnel, use analyze_funnel_dropoff.

Which analytics events fired and when. Per-event totals plus a day-by-day timeline from Firebase, GA4, or PostHog — the timeline the Firebase overview page does not show.

Where users actually drop out of your funnel. Reads per-stage user counts, conversion rates, and abandonment from Firebase, GA4, or PostHog, then explains the biggest leak in plain language.

Show which events your product already sends and propose a funnel built from them. Also lists standard journey events you are missing and why each matters.

Wire each funnel stage to the analytics event that marks it, so drop-off can be measured for real.

Diagnose why your analytics shows totals but not journeys. Checks key events, custom events, registered parameters, user identity, and data streams — with the exact fix for each gap.

Every funnel with its measurement status — how many stages are wired to events, which platform, and when it was last measured.

Query live data from any platform — Slack channels, Stripe customers, GitHub issues, Notion pages, Linear projects, Calendly events, and more. Supports platform-specific actions.

Change which site/repo/project a connected integration tracks, or which cloud region it reads from (PostHog US vs EU). Omit propertyId to auto-discover available options. Auto-syncs after switching.

Disconnect a connected integration.

SEO data from Google Search Console, SEMrush, or Ahrefs. Reports on the website of the scoped product, so several products each get their own numbers without switching a shared setting. Also shows your target keywords from your product definition.

How your site shows up in Google AI Overviews and AI Mode, plus the full Search Console search-appearance breakdown. Google exposes generative AI data in its web interface only, so this reports what the API does have and tells you how to import the rest.

What happened in your dev tools that marketing should know about — shipped features, releases, support feedback, competitor mentions — classified by marketing relevance rather than handed over as a changelog.

Pull the latest issues, releases and feedback from a connected dev tool and re-classify them.

A weekly gap analysis between what you shipped and what you announced. Most developer-founders ship far more than they say out loud; this names the releases that never got a post, and the testimonial candidates sitting in support threads.

Your configured webhook endpoints for agent and automation platforms, with their status.

Push your current marketing context to a webhook endpoint on demand — n8n, Make, Zapier, or your own service.

Store AI search impressions exported from the Generative AI performance report in Search Console, so they sit alongside every other metric. Impressions only — that report publishes no clicks, CTR, position, or queries.

Post a message to a Slack channel. Auto-resolves channel names.

List available Slack channels with member counts and topics. Call before send_slack_message.

Upload a file or text snippet to Slack.

Add an emoji reaction to a Slack message.

Post to a Microsoft Teams channel. Includes attribution.

List Teams teams and channels.

Create a new channel in a Teams team.

Read recent messages from a Teams channel.

Post to a Discord channel via webhook.

Send email via your own platform — SendGrid, AWS SES, or Resend.

Publish or draft a blog post on your WordPress site.

Publish content calendar items to WordPress, Ghost, or Notion.

Create a page or database entry in Notion.

Post to a subreddit.

Post to your LinkedIn feed.

Post a tweet.

Create a Google Doc.

Update cells in a Google Sheet.

What the connected account can do right now — profile and follower counts, and whether public pin search has been approved yet. Probed live, because a declined scope and a pending beta look identical from config.

List boards with pin counts and privacy. Every pin needs a board id, so this usually comes first.

Create a board. Secret boards are the right home for research and work in progress.

Rename a board, change its description, or switch it between public and secret.

Delete a board and every pin on it. Permanent — Pinterest has no undo.

Search the account's boards by name and description, to find where something was filed.

List pins across the account or on one board, optionally with 90-day and lifetime metrics — which turns an inventory into "which of these worked".

Read one pin in full — copy, destination link, image URLs at every size, dominant colour, and any AI disclosure its creator made.

Publish a pin from an image URL — the last step after commission_creative. Carries alt text and an AI disclosure Pinterest reads for its own labelling.

Edit a live pin or move it between boards. Fixing the words keeps the pin's saves and age; republishing loses both.

Delete a pin. Permanent, and its accumulated saves and impressions go with it.

Save an existing pin — including someone else's — onto a board, with the original creator's attribution intact. This is how an inspiration board gets built without republishing anybody's work as your own.

Per-pin impressions, pin clicks, outbound clicks, saves and save rate, as a summary, lifetime totals and a daily series.

Search Pinterest for pins and get back real images to open. Public search across all of Pinterest where the app has been approved for it; the account's own saved pins otherwise — and the response always says which corpus answered.

Top trending search keywords by region — growing, monthly, yearly or seasonal — with growth rates. Pinterest search runs months ahead of the event, so seasonal trends tell you when to publish, not just what.

Check if your calendar is connected. Built-in calendar is always available; Google Calendar is optional.

Add an event to your calendar — interviews, outreach blocks, content publishing reminders.

Show upcoming events.

Remove a calendar event.

Schedule a customer interview with prep notes and optional Calendly/Zoom link.

Schedule content publishing on your calendar.

Show your Calendly booking page types.

Create a new Calendly booking page. Free plans allow 1 event type — use create_calendly_scheduling_link instead.

Generate a booking link for an existing event type. Free plans: omit maxUses. Paid plans support limited-use links.

Show upcoming or past Calendly events.

Cancel a scheduled Calendly event.

Create a Zoom meeting.

Create an event on the built-in calendar and optionally Google Calendar.

Schedule a recurring job — sync integrations hourly, send weekly report emails, post to Slack every Monday, or fire webhooks on a cron schedule.

Show all scheduled automations and their last run status.

Pause, resume, or update an automation job.

Execute an automation immediately, regardless of schedule.

Fire a platform event to trigger all event-based automations listening for it.

Generate a webhook URL for Zapier — receive notifications from 6,000+ apps as events.

Everything needed to write a positioning statement, gathered in one call — your product definition, what interviews actually said, competitor claims, personas and category benchmarks.

Score your positioning 0-100 on specificity, grounding in research, differentiation from competitors, and whether it uses your customers own words. Returns what to change, not just the number.

Turn a persona into a system prompt, so your AI tool answers AS your buyer — same objections, same budget, same scepticism. Rehearse the conversation before you spend a real one.

Score a pitch, cold email or landing page from that buyer point of view. Would they take the meeting, what would they object to, what landed. 0-100 with specific fixes.

The data behind a price — competitor pricing, the budget range on your persona, budget signals from interviews, and benchmarks for your category. Priced from evidence rather than from your hosting bill.

Score an offer 0-100 — price against the persona budget, price against competitors, guarantee strength, bonus value and completeness.

Write your own account of a day — what you did, what it achieved, what blocked you, your energy. Activity counts and achievements are captured from the platform automatically and this never overwrites them: the two halves stay separate on purpose. Defaults to today.

Read the diary back — per-day entries, day numbers, totals, your current streak, and which milestone reports you have reached. Empty days are shown rather than skipped, because a gap is information.

A milestone report over 7, 30, 180 or 365 days — what you did, what it produced, and how the period compares. Returns the report data and a link to download it as a PDF.

The milestones you have earned, with their tier — Bronze through Diamond — and when each was unlocked.

How close you are to the ones you have not earned yet, so the next milestone is a target rather than a surprise.

Run detection now and award anything newly earned. Runs automatically as you work; this is for when you want to see it immediately.

A public link, an SVG badge, a social card and embed code for one earned achievement. Building in public is easier when the proof is a URL.

Which providers are supported, which you have connected, and what each can generate. Answers 'why did I get a prompt instead of an image'.

Landing page sections, cold emails, ad variants, social posts — written in your brand voice, from your product and persona data. Returns a job to collect, or a ready-to-paste prompt when there is nothing connected to wait for. Both are a result: a request never ends with nothing.

Ad creative, social graphics, hero images — with dimensions, style and negative constraints already specified. Returns a job to collect, or a fully-specified prompt you can take anywhere.

Turn a script into spoken audio for demo voiceovers, ad reads and podcast intros. Returns a job to collect, or the script formatted with delivery direction.

Collect a queued generation. Poll every five seconds while it is running. Generation never happens inside the request that asked for it, so a slow provider cannot hold up anything else.

The benchmark thresholds, grading criteria, formulas and required integrations. Call this BEFORE computing a report, so every metric is graded against a stated bar rather than an impression.

Publish a report with the computed scores. Only the scores and the working are stored — never the raw data read out of GitHub, Stripe, GA4 or Sentry to produce them.

Your plan, when it renews or ends, payment status, recent invoices, and whether a voucher is providing it.

A Stripe Checkout link for a paid plan — buy now, or start 30 days free before the first charge. Returns a URL you open yourself; no tool can complete a purchase on your behalf.

What a code would give you, WITHOUT applying it — the plan, how long it lasts, when it ends, and which plan your account returns to afterwards.

Apply a voucher. Access is temporary and reverts to the plan you were on before, never to Free by default. Refused on an account with a paid subscription, because it would overwrite a plan you are paying for.

Issue a new API key and revoke the current one, with per-client instructions for updating your configuration. The new key is shown once.

What the platform is permitted to do with your data, purpose by purpose, with the legal basis for each.

Turn on a processing purpose — coaching, analytics, AI generation. Takes effect immediately.

Turn a purpose off and erase what it produced. Reports exactly what stopped, what was deleted, and what was kept — revoking analysis of your diary deletes the reports, never the entries you wrote.

Everything held about you, why, for how long, and who else sees it — rendered from the same inventory that drives erasure and export, so it cannot disagree with what the code does.

Find journey records that are not linked to any product. Unassigned records are skipped by product-scoped coaching, weekly plans, and journey health.

Link an existing record to a product. Fixes orphaned data and moves records between products.

Delete any record by type and ID — positioning, funnel, experiment, competitor, etc.

Delete a product/project.

Delete a goal.

Delete an automation.

Export everything as a complete launch package — products, research, positioning, personas, offers, GTM, funnels, competitors, experiments, and content calendar. JSON or Markdown.