● Platform
Anyone can chart an ads API. Matching every krone to the click that caused it, judging it after COGS and lifetime value, and then letting an agent safely change your campaigns — that is five hard systems working together.
Connectors
OAuth · webhooks
Attribution
click-id → order
Profit engine
COGS · LTV
AI reasoning
verdicts + evidence
MCP actions
you approve · we push
OAuth apps reviewed by Meta and Google, incremental sync with backfill, and webhooks for anything that can't wait for a poll. No pixels to paste, no scraping, no CSV exports.
A first-party script captures click ids at landing, and orders are matched back deterministically — not modelled, not guessed. What can't be matched is labelled unattributed rather than quietly assigned to whoever shouts loudest.
Product-level COGS from Shopify, subscription economics from Stripe, refunds and fees deducted. Every campaign is judged on money you keep — and on marginal ROAS, so scaling advice is a model, not a hunch.
The AI layer runs on your resolved data, not on screenshots of it. Every verdict cites the numbers behind it, so you can disagree with the reasoning instead of trusting a black box.
Meta and Google are exposed to the agent as typed, permissioned MCP tools. It can propose a budget change or pause a keyword — but nothing leaves the building without your approval, and everything is logged and revertible.
What MCP actually means
MCP — the Model Context Protocol — is the open standard for giving AI models typed, permissioned access to real systems. Most “AI marketing tools” can only write you a summary. Kepra exposes Meta and Google as callable tools, so the agent can propose a real change — and the approval gate makes that safe.
Chatbots
Describe the problem back to you.
Dashboards
Show the number and wait.
Kepra
Proposes the fix, executes it on approval.
Typed, scoped, logged. Reads are free; every write waits for you.
meta.campaign.pause(campaign_id)approval requiredmeta.adset.set_budget(adset_id, amount)approval requiredmeta.audience.create(spec)approval requiredgoogle.keyword.pause(keyword_id)approval requiredgoogle.campaign.set_bid(id, target_roas)approval requiredkepra.profit.query(range, dimension)read-onlykepra.leak.list(threshold)read-onlyEvery write is proposed by the agent, approved by you, executed over MCP, and written to an immutable audit log — with a one-click revert.
Every write is a proposal until you press approve. Autopilot is a setting you may never turn on — and we ship it off.
Data lives in the EU (Ireland) with row-level security: an org can only ever read its own rows, enforced by the database, not by application code.
Who proposed it, what evidence it cited, who approved it, what the platform answered — written down, permanently.