Keiki connects the same agent to Slack, WhatsApp, and email

Orchid is launching Keiki, a platform for creating an agent with company knowledge, memory, and tools, then connecting it to major messaging services.

Orchid is launching Keiki, a platform designed to deploy conversational agents across the channels customers already use. Instead of adding another chatbot to a website, the company proposes placing the same agent inside email, Slack, WhatsApp, Telegram, SMS, or Apple messaging conversations.

Configuration begins with information specific to the organization. Users can add products, procedures, documents, writing guidelines, and operating rules. Keiki then uses this material to build an agent capable of responding in a consistent voice across different channels.

The agent also includes memory, allowing it to retain certain information throughout an exchange, and can be granted access to a browser or external tools. Orchid demonstrates the creation of an email support agent using company knowledge, a model selected by the user, and conversation history.

Keiki is not limited to answering questions from documentation. Connected tools can allow it to recommend products, qualify leads, schedule appointments, initiate payments, or hand a conversation over to a person. Sensitive actions can require human approval, while configurable boundaries determine what the agent may complete on its own.

In theory, this approach maintains the same personality, information, and tools across multiple platforms. An email request could therefore be handled according to the same rules as a question received through Slack or WhatsApp. Keiki has not yet explained, however, how user memory might be reconciled across separate identities on different channels.

The company also emphasizes agent observability. Its interface centralizes conversations, execution traces, tool calls, and outcomes. This information can help teams investigate an error, understand why an action was triggered, or identify responses that require updated instructions.

Orchid presents these traces as a source of data for gradually improving agents through reinforcement learning. It has not provided technical details about the training process, the models involved, or how customer conversations would be selected and prepared for this purpose.

The platform includes preconfigured templates for different professions. Examples cover dental practices, real estate agencies, restaurants, retailers, healthcare services, recruiters, and independent professionals. These templates prefill some instructions and functions, but still need to be adapted to each organization’s information and operational constraints.

For a restaurant, the agent could answer questions about the menu, business hours, or reservations. In real estate, it could qualify an inquiry and schedule a showing. For a medical practice, it could process appointment requests, send reminders, or escalate a situation requiring human intervention. Several examples shown on the website, however, use fictional businesses and simulated transactions.

According to Orchid’s launch announcement, Keiki is already handling more than one million messages per month. The company has not disclosed the number of deployed agents or how that volume is distributed across channels, making the figure difficult to evaluate in context.

The initial announcement lists Slack, email, iMessage, Telegram, and WhatsApp, with Microsoft Teams expected to follow. The current Keiki website instead mentions SMS, WhatsApp, Slack, Telegram, email, and Teams. Orchid has not yet published detailed documentation explaining the technical requirements for each connection, particularly Apple’s messaging service.

Keiki also states that it is SOC 2 and HIPAA compliant. These claims are relevant to organizations handling sensitive information, but they do not remove the need to review agent settings, accessible data, conversation retention periods, and the conditions imposed by connected third-party services.

The platform is currently available through an early-access registration process. No public pricing, complete list of supported AI models, or service-level commitments have been published. Keiki therefore positions itself less as a simple chatbot builder and more as an orchestration layer designed to keep one agent consistent across multiple conversations, while giving the organization visibility into its decisions and the ability to take control when necessary.