Keiki

Companies that want to let customers interact with their product conversationally face the work of building and wiring up agents across many separate channels like iMessage, WhatsApp, SMS, Slack, Teams and email, each with its own setup. Customers often have to open a separate app to get support, check billing, change plans, or ask questions that could be answered in a text message. Support teams must manually handle routine requests like seat upgrades, invoice lookups, plan changes, refunds and call bookings that could otherwise be resolved automatically. Connecting an agent safely to a company's real systems, so it only takes actions a given customer's account allows, is also a nontrivial integration and security problem.

Keiki lets a company describe an agent in plain language, teach it their product by pointing it at their website, docs and API or MCP server, and publish it across iMessage, WhatsApp, SMS, Telegram, Slack, Teams and email from one agent. The agent gets its own number for iMessage and SMS, answers customer questions from company docs and policies, and can take actions in the product such as adding seats, switching billing plans, pausing subscriptions or pulling invoices, asking for approval before charging anything if configured to do so. Built-in capabilities include web browsing, a code execution sandbox, persistent memory of past conversations with each customer, scheduled recurring tasks, and an agent wallet with spending limits for tasks that require payment. Every action is checked against the customer's own account permissions, conversations can be read or taken over by a human at any time, and the platform supports multiple underlying models (Claude, GPT, Gemini) with SOC 2 and HIPAA compliance. No coding is required to build an agent via the dashboard or templates, though developers can extend it with an SDK and API.