The autonomous growth operating system
Give Xyranik a goal.It runs the growth loop.
Tell it what you want to grow. It finds what’s worth doing in your own sales, prepares the emails, posts and ad changes, and carries out what you approve. Then it shows what each one really brought in.
14-day evaluation · no card · start before you connect anything
Why it’s different
Other software waits for instructions. Xyranik starts from a goal.
Marketing tools generate, suggest and automate, then leave the deciding and the measuring to you. Xyranik understands the business, decides, orchestrates the work, carries it out, measures it and learns. It asks you at the one point that matters.
Traditional marketing software
Every decision is yours, and so is working out what happened.
- 1You decide what to doYou
- 2You set it up, channel by channelYou
- 3It sends what it was told to sendSoftware
- 4You export a reportYou
- 5You work out what it meant, and what to try nextYou
Xyranik
You set the goal and approve the work. It does the rest, and shows its working.
- 1You say what you want to growYou
- 2It learns the business and finds what is worth doingXyranik
- 3It plans the work, citing your own recordsXyranik
- 4You approve it, or let it act within limits you setYou
- 5It runs the work through agents and workflowsXyranik
- 6It measures the result, learns, and recommends the next stepXyranik
- Simple on the surface
- A goal in plain words, and one Inbox for everything that needs you.
- Powerful underneath
- Agents, workflows, channels and a measurement ledger, working from one plan.
- Intelligent throughout
- Every recommendation starts from your own records and says which ones.
- Trustworthy by design
- It acts only within the authority you give it, and every action is on the record.
How it works
One loop, from a goal to a measured result
You give it a goal. Xyranik works through 7 steps and stops for you at the one that matters: before anything goes out.
- 1
Step 1 · Understand
It learns the business
Xyranik reads your website into facts, and each fact quotes the page it came from. The line items of your orders show what actually sells. Where you say otherwise, your word wins.
Where you see it: Your business
- 2
Step 2 · Decide
It finds what is worth doing
The opportunity engine works through your own records: regulars going quiet, a channel slipping behind its usual pace. Each opportunity comes with the measured gap behind it, or it is not raised at all.
Where you see it: Home
- 3
Step 3 · Plan
It writes the strategy
The AI CMO turns your goal into a mission it can measure, and writes a strategy from an evidence pack. Every claim cites its evidence, and every figure is checked against it before you see it.
Where you see it: Goals
Your step · Approve
You decide
Proposed changes, sends waiting for approval and strategies all wait in one Inbox. Each says what will happen and why, and is decided right there. Nothing goes out until it is, unless you have given that kind of action more room.
Where you see it: Inbox
- 4
Step 4 · Execute
It does the work
Agents and workflows carry out what was approved: email, SMS and WhatsApp, social posts and ad budgets. Each agent may make only its own kinds of change, and every run is recorded.
Where you see it: Agents and Automations
- 5
Step 5 · Measure
It counts the real result
Sales land in one order ledger, each counted once and traced by tracked links and discount codes. Hold-back groups show what an email actually added, and each month closes into a statement of what Xyranik cost against what it brought in.
Where you see it: Results
- 6
Step 6 · Learn
It keeps score on itself
Every recommendation went into the decision ledger with its prediction before the outcome was known. Afterwards it is scored against what happened, so the track record is there to read, misses included.
Where you see it: Decisions
- 7
Step 7 · Improve
It recommends the next step
What held up and what missed become evidence for the next decision. Where its record falls short, Xyranik lowers its own authority for that kind of action. Only a person can raise it again.
Where you see it: Home
Then the loop turns again: each result becomes evidence for the next decision.
Each turn is written down before the outcome is known, so the record can be read back against what happened.
What it learns before it decidesWhere a prediction is scoredEvery week
One email that tells you what happened
Every Monday, the week just ended in one email: what needs you, how your goals are going, what sold, what Xyranik did and how its predictions held up. Your team can have it too.
What needs you comes first
Anything waiting for a decision leads the email, with a link to decide it in Xyranik.
Only what was counted
Sales net of refunds, per currency and never added across currencies, with how much was traced to your marketing. A part it could not read is named, not left out.
Its record, misses included
The predictions that reached their date that week, right or worse, and its whole record in one sentence.
No news, no email
A week with nothing in it is not sent, and one link in every email turns it off.
Nothing in the briefing is generic advice: every line is drawn from this workspace’s own records, and where there is nothing to say it says so.
How it decides what is worth doingAI CMO
A strategy you can check, line by line
Say what you want to grow, in plain words. Xyranik turns it into a mission it can measure, then writes the strategy from an evidence pack drawn from your own business.
- 1
Your goal becomes a mission
Read from your own words and measured against a baseline from your own records, so progress is a number you can check.
- 2
Every claim cites its evidence
Each line of the strategy points to the evidence behind it, and every figure in it is checked against that evidence before you see it.
- 3
Approve it, and the work is handed out
Workstreams go to the agents, and each decision goes into the ledger with its prediction.
- 4
Reviewed every day
The plan is checked daily against what is actually happening, and you are told when it drifts.
The evidence pack is assembled before any model is asked anything, and every claim is checked back against it.
How a strategy is builtWhat makes a goal measurableAI workforce
Agents defined by what they may do
Each agent is a capability with a boundary, not a persona. It works from your business’s context toward your goal, may make only its own kinds of change, and leaves a record of every run.
Every agent, and the boundary it works inside
Analyst
Works on: Your orders, channels and results
May propose: Nothing. It studies and reports.
Operator
Works on: Ad campaigns
May propose: Pause a campaign, change a daily budget, or move budget between two campaigns
Automator
Works on: Automations
May propose: Draft, turn on or pause a workflow
Retention
Works on: Customers going quiet
May propose: Draft a win-back email
Content
Works on: Your social accounts
May propose: Draft a post, held for approval
Responder
Works on: Email people send you
May propose: Draft a reply to one email, held for approval
Visibility
Works on: How AI assistants answer questions about you
May propose: Nothing. It studies and reports.
A proposal of any other kind is refused by name, and no agent can widen its own boundary. What it may do without asking is set by you, one kind of action at a time.
An agent is a set of boundaries before it is a capability, and both are yours to set.
Every agent, and the boundary it works insideRules you write in your own wordsAutomations
Describe it. Check the dry run. Turn it on. See what it did.
Automations are written in plain words and rehearsed before they reach a single customer.
- 1
Describe it
Xyranik drafts the workflow from your words, using only steps it really has.
- 2
Check the dry run
It says what it would do without doing any of it. A step it cannot match is named, never guessed.
- 3
Turn it on
Each time it is turned on, a version is frozen and the change goes into the audit log.
- 4
See what it did
Each run shows every step it took, and why it stopped.
A wait here is durable: it survives a restart rather than holding a process open.
What an automation can doHow work is carried outTrust and control
Autonomy without a blank cheque
Every agent has an identity, permissions and limits, and every action one asks for is decided before it happens — by rules you wrote, not by the model asking. Autonomy is set for each kind of action, not all at once. You raise it. Xyranik can only lower its own, automatically, when its record falls short.
The autonomy ladder
Example: pausing an ad campaign
- OffNothing runs on its own.
- ObserveIt works out what it would do and writes it down, without doing any of it.
- ProposeIt proposes work, with what it expects each piece to achieve.
- ApproveIt prepares real changes and waits for your approval.Set by you
- ActIt carries out its own changes, inside the limits you set.
Raised only by the owner. Once 10 of its changes of one kind have been measured and 8 in 10 were right, Xyranik may ask for room to make that kind on its own, up to a weekly limit.
Lowered by Xyranik, a step at a time, when its predictions keep missing.
Every action leaves a trail
- 1Pending
Waiting in the Inbox, saying what will happen and why.
- 2Approved
By a named person, or within the level you set.
- 3Executed
Marked done only once the change went through on the platform itself.
- 4Measured
Scored against the prediction written before it ran.
Before anything runs, the prediction is written down, so every action can later be shown to have been right or wrong.
Every agent is somebody
Each has its own identity, its own permissions and its own limits — so “pause that one” is something you can actually do, and one stopped agent does not stop the rest.
Your rules, in your words
Ask me before an ad budget goes up by more than 20%. Never let an agent take more than 100 people’s details. Rules are checked before anything happens, and a request that leaves out the figure a rule is about is refused, not waved through.
One switch stops all of them
It refuses everything from that moment: new runs, changes waiting to be carried out, and anything an agent asks. A run already thinking stops at its next step. Your own people carry on as normal.
Approvals
Sends, changes and strategies wait for a person, in one Inbox.
An append-only audit log
Who did what: a person, an API token, the system or an agent. Nothing edits or deletes an entry.
Roles
Owner, admin, member and viewer, checked on the server for every change.
Single sign-on and SCIM
Sign in through your identity provider over OpenID Connect; add and remove people over SCIM.
Encrypted and kept apart
Keys and tokens for your connected accounts are encrypted before they are stored, and each workspace’s records are kept apart.
Your data, in one place
See what Xyranik holds and where it came from, download it, erase one person when they ask, and choose how long form answers are kept.
No security certification is held yet, and none is claimed.See what is in place
What actually worked
Attributed is not the same as caused
Most software shows you what happened after a campaign and lets you read it as a result. Xyranik traces each order once, keeps what a send added separate from what merely followed it, and says so plainly when it cannot tell.
- Each order is traced once, in a fixed order
- A discount code, then the last thing they clicked, then the first — and then an explicit “we cannot tell”. One order is never counted for two channels, which is how a business ends up with four tools each claiming the same sale.
- What a send added, not what followed it
- Hold back a share of the audience and the difference between the two groups is what the send actually caused. Attributed and incremental are shown as different figures throughout, because acting on one while meaning the other is how money goes into the wrong channel.
- A closed month stays closed
- Months settle into a statement that is not revised afterwards, so a figure you acted on in April still reads in October the way it read then. A history that quietly improves is not a record.
- An unknown figure stays empty
- Where there is not enough evidence, you get an em dash and a sentence saying what would fill it — never a plausible zero. “We cannot tell you” and “the answer is nothing” lead a business to opposite decisions.
Start small
Start without handing over your data
Xyranik is useful from the first conversation and earns more access as it proves itself. You decide what it sees, and when.
- 01
Start with context
Tell it what you sell and what you want to grow. It drafts a starting plan from what you say.
Works fromWhat you tell it
- 02
Connect what you choose
Your store (Shopify, WooCommerce or Stripe), your ad or social accounts, or a file of past orders, one at a time. Each is optional.
Works fromWhat you connect
- 03
Evidence
Recommendations now cite your own records, and results are measured in your own ledger.
Works fromYour own records
- 04
Controlled execution
Work goes out only through approval, at the level you set for each kind of action.
Works fromYour approval
- 05
Earned autonomy
Where its predictions hold up, you can give it room to act. Where they don’t, it steps back on its own.
Works fromIts track record
Connecting is a decision you make one source at a time, and what each one adds is said before you make it.
How one workspace is kept from anotherWhat connecting a source addsPlatform
Built to become infrastructure
Marketing is where Xyranik starts. Underneath is a layer that understands a business, decides, carries out the work and governs it, and developers build on it over the same API and the same gateway.
What you build on
Available now- A versioned public API
- Version 1, at /api/v1. Read goals, the decision ledger, signals, opportunities, experiments, customers, revenue and usage. Add contacts, report events from your own systems to start workflows, and propose changes, which wait for a person exactly as an agent’s do.
- Scoped tokens, live and test
- A token carries only the permissions it is given. Test tokens read real data and only simulate changes.
- Safe to retry
- Send an Idempotency-Key with a write. The same request again within 24 hours gets the first answer back, and nothing is done twice.
- An OpenAPI description
- Generated from the same list of endpoints the API serves, so the two cannot disagree.
- Signed webhooks
- Events delivered to your own systems, each signed so you can check it came from Xyranik.
- A trust layer under every agent
- Each agent has an identity of its own, the permissions it was given, and limits. Every action one asks for is decided before it happens — against what that agent may do, the data it may use, how risky the request is and the rules you wrote — and written down with the answer, refusals included. Xyranik’s own agents go through it, which is the only reason to believe it works.
- The same trust layer, for the agents you build
- An agent you built asks the same question over the API — may this agent do this, here, now? — on a credential of its own, and is held to the same permissions, the same limits and the same record. The answer says allowed, denied, or that a person here has to decide first, which is not a yes. Revoking one agent’s credential stops that agent and nothing else.
Read the decision ledger with a test token
Example# Every recommendation, what it predicted, and what happened
curl https://xyranik.com/api/v1/decisions \
-H "Authorization: Bearer xyr_test_…"Your own agents ask the same gateway ours do, on a credential of their own, and are held to the same record.
All three layers, in orderBuild on the same gatewayWho it’s for
The same loop, pointed at different work
Xyranik starts from what a business already has, whether that is a store, a website, a team or an audience.
Online sellers
Connect Shopify, WooCommerce or Stripe. Xyranik notices regulars going quiet and proposes a win-back, measured against a hold-back group, and judges your ads by your own orders rather than the platform’s report.
Service businesses
No store needed. Start from your website and your goal, bring in past sales from a file, and follow up every enquiry from a hosted form automatically.
Marketing teams
One Inbox for approvals, roles for who may change what, workflows with versions, and an audit log of every change, by a person or an agent.
Creators
Hooks, scripts and captions in your own voice for 7 formats, a content calendar, publishing to Instagram, LinkedIn and X, and paid work from brands on Xyranik.
Larger organisations
Single sign-on over OpenID Connect, SCIM provisioning, an append-only audit log, data retention set per workspace, and a versioned API.
Each of these has a page saying what Xyranik does for that reader — and what it does not.
Find the one that fitsFor creators
From an idea to a script, in your own voice
The Studio turns an idea, a trend or a brand brief into hooks, a timed script, a shot list, a caption and hashtags, written the way you talk, for 7 formats. It designs the carousel slides and the cover too. Then it goes on your content calendar.
- Instagram Reel(published from Xyranik)
- TikTok video
- YouTube Short
- YouTube video
- Instagram carousel(published from Xyranik)
- LinkedIn post(published from Xyranik)
- X post(published from Xyranik)
Published from Xyranik to Instagram, LinkedIn and X. TikTok and YouTube pieces are posted from their own apps.
Brand work, in the same place
Brands look for creators on Xyranik, and you find paid work beside what you make. The brand pays you directly; Xyranik keeps the record, from the agreed fee to your confirming it arrived.
- A profile and portfolio brands find only when you turn it on
- Campaigns ranked by how well they fit you, and why
- Terms agreed by both sides before any work starts
- Work sent for review from the Studio, with the rounds of changes agreed
Pricing
Evaluate it for 14 days, then choose a plan
Every plan has the same loop. They differ in volume, seats and support. AI work is counted in Xyranik credits.
Free
14 days to see whether this works for you. No card.
- 250 contacts
- 1,000 emails
- No SMS
- 10 social posts
- 500 Xyranik credits
- 1 active automation
- 1 team member
- 3 rules for your agents
Basic
Solo operators and single-brand businesses
$29a month
or $278 a year (20% less)
14-day trial
- 2,500 contacts
- 10,000 emails a month
- 100 SMS a month
- 100 social posts a month
- 15,000 Xyranik credits a month
- 10 active automations
- 2 team members
- 10 rules for your agents
- A/B tests
Pro
Marketing teams running several channels at once
$79a month
or $758 a year (20% less)
14-day trial
- 15,000 contacts
- 30,000 emails a month
- 250 SMS a month
- 500 social posts a month
- 40,000 Xyranik credits a month
- 50 active automations
- 5 team members
- 50 rules for your agents
- A/B tests
- API and webhooks
- Priority support
Advanced
Agencies and multi-brand marketing organisations
$199a month
or $1,910 a year (20% less)
- 60,000 contacts
- 70,000 emails a month
- 750 SMS a month
- 2,000 social posts a month
- 120,000 Xyranik credits a month
- 200 active automations
- 25 team members
- 200 rules for your agents
- A/B tests
- API and webhooks
- Priority support
Enterprise
Large organisations with a named account team
Custom
Priced with your account team, around what you need.
- 200,000 contacts
- 150,000 emails a month
- 1,500 SMS a month
- 10,000 social posts a month
- 350,000 Xyranik credits a month
- Unlimited active automations
- Unlimited team members
- 1,000 rules for your agents
- A/B tests
- API and webhooks
- Dedicated support
Compare every allowance, and read the questions on the pricing page.
Give Xyranik a goal, and see what it would do.
It learns the business, proposes a plan, and waits for your approval before anything goes out.
14-day evaluation · no card · start before you connect anything