What is Akii?
Akii gives agencies one approved Commerce Graph for each client. It holds services, offers, prices, policies, locations, sources, and customer actions. Answer QA checks AI answers and buyer paths against those facts.
Updated 2026-07-08
Search setup, plans, safety, protocols, Answer QA, and customer paths. The short version is usually here.
Plain answers about what Akii is, who it helps, and how to start.
Akii gives agencies one approved Commerce Graph for each client. It holds services, offers, prices, policies, locations, sources, and customer actions. Answer QA checks AI answers and buyer paths against those facts.
Akii fixes scattered client facts. It gives AI systems one approved way to read what a client sells, what rules apply, and where a customer should go next.
No. Akii is the client data and action layer behind AI answers. It helps chatbots and agents understand the client, but it is not a chatbot.
No. Akii prepares approved client data and action paths. It does not control third-party AI models, rankings, inclusion, or recommendations.
Akii includes Commerce Graph, Agent Gateway, Action Registry, Protocol Compatibility, Answer QA, Monitoring, Fix Queue, and Sources.
Create an account. Pick the plan size you need. Add your first client brand. Build its Commerce Graph. Map customer actions. Publish Agent Gateway. Run Answer QA.
Which businesses, teams, and partners Akii is built for.
Akii fits businesses that get customers online. It helps with buying and booking. It also helps with calls, messages, quotes, forms, and deposits.
Yes. Akii helps product businesses set clear offers and policies. It also maps variants, bundles, stock status, and checkout handoffs so AI systems can read them.
Yes. Akii supports service businesses. It can map bookings, quotes, calls, and WhatsApp paths. It can also map deposits, forms, policies, locations, and human handoffs.
Yes. Akii can model locations, hours, service areas, contact paths, policies, reviews, listings, and branch-level sources.
Yes. Akii helps quote-based and inquiry-led B2B companies map offers, proof, policies, forms, partner routes, and sales handoffs.
Yes. Agencies can onboard client brands, build graphs, map action paths, run QA, manage fixes, and package reports without building the data layer themselves.
Commerce Graph, Agent Gateway, Action Registry, and core workflows.
A Commerce Graph is the approved profile of a client business. It can hold services, offers, prices, and policies. It can also hold locations, sources, and customer actions.
A client brand is one service-business client in Akii. It keeps data, sources, customer actions, issues, fixes, and settings together.
Sources prove client facts. They can be websites, listings, menus, PDFs, policy pages, booking systems, and reviews.
Draft data is still being prepared. Approved data has been reviewed. Published data is the public output AI systems and partners can read.
Agent Gateway shares approved client data with AI tools and partners. It helps them read offers, policies, locations, and next steps.
Action Registry maps each offer to the right next step. That can be buy, book, inquire, request a quote, call, WhatsApp, pay a deposit, submit a form, or check out.
Yes. One offer can have different paths by market, device, language, provider, buyer intent, or fallback rule.
Protocol Compatibility checks if data, policies, and handoffs are ready for AI commerce.
Answer QA checks AI answers against approved client data. It also checks public sources and customer paths.
How Akii shares approved client data with AI tools, partners, and customers.
Akii publishes approved public data. This can include facts, locations, offers, policies, and next steps.
No. AI-facing output should only use approved public data. Private notes, unapproved work, account settings, and review notes stay out.
Read-only access lets approved AI tools read public data without changing it. It helps them understand offers, policies, locations, and next steps.
No. Akii AI access is read-only. It does not create bookings, submit forms, process payments, edit client data, or run customer actions.
Yes. Akii can turn approved client facts into structured data. This helps search engines and AI tools read offers, policies, and customer actions.
Yes. Akii can prepare scoped exports for approved partners. Partner access can be separate from public AI-facing output.
Yes. Akii can provide a plain text view of approved public client data. AI systems can use it instead of guessing from scattered pages.
How Akii checks client data for AI commerce standards and safe claims.
No. Readiness means Akii checks client data and handoff paths for a protocol-style flow. Certification comes from a separate outside program.
ACP readiness checks seller identity. It also checks offers and policies. It checks price rules, deposits, stock status, and action paths too.
Akii supports UCP-style client profiles and readiness proof. It does not claim Google certification. It does not run checkout.
Akii can expose A2A discovery through an Agent Card. Where set up, it can support safe client reads or handoffs.
AP2 readiness checks client-side payment handoff data. It also checks policy rules, support paths, and action limits. It does not process payments.
They are structured UI payload exports for approved offers and actions. Akii emits safe data payloads, not arbitrary code.
AI commerce protocols need clear client data. They also need policies, approved outputs, and handoff rules. Akii prepares that client layer.
No. Akii does not claim a formal partnership with OpenAI, Stripe, Google, or protocol groups. Akii also does not claim certification unless it is verified.
Checks, changes, issues, fixes, and retests.
Answer QA tests AI answers against approved client data. It also checks public sources, customer paths, and published output.
Akii records setup, proof, confidence, and retest history. It separates confirmed issues from likely issues. It flags findings that need review.
An issue is a tracked problem. It can show severity, proof, the affected item, the recommended fix, and status.
Fix Queue turns issues into work. Each item can show an owner, proof, next action, priority, status, and retest needs.
Akii checks if client data is fresh. It also tracks source changes, AI answer quality, readiness, issues, and customer paths.
A changed page or setting is not enough. Retesting checks whether the source, AI answer, readiness check, or customer path now works.
Plan limits, client brands, upgrades, and Partner Platform.
Akii offers Agency Launch: $99/month for 1 client brand; Agency Portfolio: $267/month for 3 client brands; Agency Scale: $799/month for 10 client brands; Partner Platform: custom client brand count.
Every paid plan includes Commerce Graph, Agent Gateway, Action Registry, Protocol Compatibility, ACP, UCP, MCP, A2A, AP2, and agent UI readiness where implemented, Answer QA, Monitoring, Fix Queue, Sources, Approved public outputs. Plans differ by client-brand count.
A client brand is one service-business client. It can include its own data, sources, customer actions, issues, fixes, and settings.
Yes. Self-serve plans are built around client brand count. Partner Platform is available for custom client-brand counts.
Choose Partner Platform when you need a custom client-brand count. It also fits teams that need portfolio planning, partner work, or sales help before setup.
What is public, what stays private, plus payments, bookings, roles, and limits.
No. Akii maps and checks booking paths. Booking completion stays with the client or the approved booking provider.
No. Akii prepares client-side data and payment handoff readiness. It does not store cards, vault tokens, issue mandates, or run payments.
Only approved public data should be shown. This can include facts, offers, locations, policies, paths, and summaries.
Private notes and unapproved work should stay out of public outputs. So should review notes. Account settings, owner details, and partner-only data should stay private.
Akii keeps roles separate. Each team can get the access it needs.
No. Public outputs should never include secrets or private credentials. They should not include privileged account data or unapproved private notes.