Client
A client is an AI tool, agent, or application connected to a workspace's context repository through an API key or via MCP.
Category
Product feature
See all terms
What is a client in Qontext?
A client is any tool, agent, or application that connects to a workspace's context repository to read, search, or write context, over the Model Context Protocol or the API. Company chats connect as clients, alongside coding agents like Codex and Cursor, workflow builders like n8n, Make, and Zapier, and any custom integration built against the MCP server or the API.
Why do clients matter?
A workspace connects several separate tools to one context repository at once: a chat assistant, a coding agent, an automation workflow. Each needs its own scoped access and its own audit trail, so a security team can point to exactly which application read or changed a file instead of treating every request as an anonymous call against an API key. Without treating each connection as its own client, access management would have nothing to scope permissions to beyond the key itself, and retrieval logs could not attribute an unexpected output to the tool that produced it.
How do clients work?
A client's access runs through its API key. An owner can let the key inherit their own read and edit access automatically, staying synced with their permissions, or manage its access independently per file and folder from the key's Access controls tab. The key is capped at can read or can edit and never reaches full access. To connect a client over MCP, point it at Qontext's hosted server at https://api.qontext.ai/mcp and complete the authentication flow, after which the client gets read, search, and write operations bounded by that key's access controls. Every retrieval a client makes gets recorded and becomes auditable: opening a client's entry in the Clients section, then its Retrieval logs tab, shows what that client retrieved and when.