Build reusable agents with knowledge tools and multichannel deployment
ChatBotKit is a platform for creating AI agents that answer questions and perform work through connected tools. It supports visual configuration alongside APIs and SDKs, so a team can start with a prepared solution or build a more customized application. The platform organizes an agent's instructions, knowledge and abilities into reusable components rather than requiring a new integration for every conversation interface.
A blueprint can combine agents, datasets, skillsets and deployment settings. Website crawling and document inputs provide searchable knowledge, while skillsets describe actions such as retrieving records or interacting with another service. The model catalog and provider-key options allow teams to choose different underlying models. This flexibility is useful, but changing a model should still trigger testing because the same instructions and tools may produce different behavior.
Agents can be embedded in a website widget or connected to messaging services such as Slack, Discord, WhatsApp and Telegram. The feature catalog also includes APIs, several language SDKs, event hooks and command-line tooling. These interfaces make it possible to integrate the platform into a product or operational workflow rather than use only a standalone chat page. Shared secrets and access controls need careful configuration when a tool can read private information or change external records.
ChatBotKit includes usage analytics, tracing and policy features to help teams observe what their agents do. Plans set resource and credit-token allowances, with higher tiers offering greater capacity and operational options. The pricing page contains older plan names in parts of its FAQ, so the current plan cards and checkout should be used when confirming an order. A useful first deployment is one agent with a limited knowledge base and a small number of approved actions. Test both answers and tool outcomes, examine traces when a task fails and keep an owner for the knowledge it uses. Reusable blueprints can simplify expansion after that initial workflow behaves reliably, but they do not remove the need to check permissions and results in each new deployment.
