Burki

What Burki is

Burki is a builder for AI phone assistants that answer calls for a business. The homepage frames the journey as "From your business facts to your first conversation": you pick a starting role, such as receptionist, appointments or customer support, add your business details, try the assistant yourself in a browser, and connect a phone line once you are happy. The reader it serves is a small business or a team that wants an AI receptionist without hiring a developer, and also the builders who put a voice agent inside their own product. The homepage says Onyx by Marc Ellis, an AI recruitment and candidate screening product, and BiteBuddy AI, which takes restaurant orders and reservations by voice, "run on Burki". The pages do not say where the company is based. They do list demo-call countries (United States, Canada, Pakistan, India, Saudi Arabia and the United Arab Emirates), and the site has Dubai and UAE pages, one in Arabic.

What you build and how

The editor is guided rather than a canvas. The homepage lays out 3 steps.

1. Make it yours. "Choose a starting role. Add your services, hours, common questions and how you want" callers helped. This is the whole configuration surface the page describes: business facts and handling instructions, not a node graph.
2. Be your first caller. You ask a question, interrupt or change your mind, "Review the transcript and adjust your draft before publishing."
3. Connect when ready. You add prepaid funds, connect an eligible number and review its readiness. The company says it can help check your carrier and routing before you switch.

The sample on the page is a plumbing-style scenario for a business called Harbor Home Services. A caller asks whether the company fixes a leaking kitchen tap, then corrects themselves to say it is the bathroom sink, and the assistant follows the correction and asks whether they want a repair or an estimate. The page labels this as excerpts from a real browser demo with fictional details, presented as a guided example.

Review is a stated strength. "Burki keeps the transcript, connected action results and usage together so your team can review the next step." The page breaks that into three parts: keep the details, including corrections and requests that still need an answer; check the outcome, meaning what connected actions returned; and understand the cost, with each call's usage and charges beside its history.

Actions go through integrations. The homepage names two. Cal.com lets the assistant "Check availability and book a confirmed appointment during the call." You connect an eligible appointment type, practice with simulated actions first, and funded calls can make a real booking after the caller confirms. HubSpot is described as finding contacts and saving contact details, with the rest of that line cut off on the page, so the exact actions are not confirmed here. A separate integrations page exists on the site.

Phone setup is carrier dependent. The FAQ says forwarding from your existing carrier to an eligible Burki-connected number "may let you keep your public number", that scheduled or conditional forwarding can send after-hours or overflow calls to the assistant, and that carrier support, destination availability and charges vary. When the assistant cannot resolve a request, you instruct it to explain what it can help with and collect the details your team needs. The company is cautious about live transfer: it "depend[s] on the voice setup and phone route", and you should contact the company to confirm before relying on a handoff.

Burki's blog explains how its comparison pages are meant to be read. One post, on Deepgram and ElevenLabs, calls itself "a documentation-based comparison" and not "a claim that either provider won a Burki benchmark". It says Burki "provides the assistant workflow around supported voice configurations", that the available provider settings determine what you can select, and that you should change the recognition or voice component only after finding out which part of the conversation needs work. Burki also publishes comparison pages against Retell, Vapi, Bland, Synthflow and RevRing.

On the chatbot side, there is no text chat widget described. Burki covers the conversational-assistant part of that category through voice only.

The evidence on pricing and limits

The pricing model is "Pay as you use it. No mandatory subscription." The worked example on the homepage reads $26.82 USD per month for 100 calls of 3 minutes, which the page puts at "About 9¢ / minute". It "Includes AI voice, an illustrative model allowance, Burki's platform charge and managed service fee", and assumes GPT Live 1 plus GPT-5.4 mini at 1,000 input and 200 output tokens per minute. Rates were "verified September 25, 2026". Phone carrier charges, number rental and management, optional actions and taxes cost extra, and per-call rounding varies. A usage estimator is offered.

Free access has two layers. The public demo runs up to 10 minutes in the browser with no signup and no phone number, using your microphone, and all three public scenarios (receptionist, appointments, customer support) offer up to 10 minutes when capacity is available. A free call from Burki to your own phone also runs up to 10 minutes. Eligible new workspaces get up to 3 minutes of browser practice, with up to 2 minutes per call and no card to start. "Trial exhaustion does not charge you automatically"; paid usage begins only when you add prepaid funds and use them.

On privacy, the public demo does not record audio, and transcripts are saved for review. Your voice goes to the AI service. Recording in your own workspace needs supported storage, consent and eligible funded usage. Demo calls cannot book, call or message anyone. The pages state no SOC 2, ISO or HIPAA claim.

Who it suits and who it does not

Burki suits a small business testing whether an AI receptionist can handle its calls: the no-signup demo and prepaid pricing keep the first experiment cheap and reversible. It also suits a developer-led team building something like the recruitment or restaurant-ordering products named above, where per-call cost visibility matters. The Cal.com booking path is a good fit for service businesses that already take appointments there.

Limits are plain. Setup is guided by instructions, so a team wanting a branching flow canvas or code-level control will not find one described. Existing numbers depend on carrier forwarding and are checked case by case, live transfer is not guaranteed, and the two named integrations are Cal.com and HubSpot only. The pages do not give a monthly plan, seat model or support tiers, so a larger team should ask about volume terms.