What is CortexHub?
CortexHub is the control plane for AI agents. Admins can create organization-owned agents, install a hosted agent as a named Slack bot, let members use personal co-workers, or bring an existing AI client through MCP. Every path uses the same governance: least-privilege access, approval for risky moves, redaction of sensitive data, and a tamper-evident audit trail.
There are three ways to deploy AI on one shared foundation:
- Org Agents. An organization-owned hosted or SDK agent runs shared work under an agent group. A hosted Org Agent can become a named Slack bot that members @mention.
- Co-workers. Each member gets one hosted AI co-worker that chats, builds a Brain of their work, runs scheduled and multi-step work, and uses that member's connected tools and files. It acts only as its owner, within the organisation's rules.
- Governed gateway for external AI. AI clients (such as a desktop AI assistant or an in-editor coding assistant) and partner backends connect to one governed endpoint and reach the tools your organisation has made available, always within policy.
Both paths run through the exact same governance before any tool executes. There is no separate, weaker path.
What changes with CortexHub
| Without CortexHub | With CortexHub |
|---|---|
| Agents hold API keys and OAuth tokens | Credentials stay in CortexHub; agents never see vendor secrets |
| Shared agents run with unclear ownership | Org Agents have explicit organization ownership, capabilities, groups, and audit |
| Employees use unmanaged chat AI | Each member gets a personal co-worker with group-bounded tool access |
| Each integration exposes a long flat tool list | One governed gateway; discovery and invoke go through meta-tools |
| Inconsistent rules per user or tool | Layered access rules that only get stricter: organisation floor, then groups, then members |
| Approvals are ad hoc | Structured approval flow that resumes the action on its own once a person decides |
| Tool calls are hard to attribute | Every call produces telemetry, audit events, and tamper-evident records |
| Sensitive data may leak into logs | Sensitive fields are redacted from stored records; secrets are always scrubbed |
Core surfaces
| Surface | URL | Who uses it |
|---|---|---|
| CortexHub Cloud | https://app.cortexhub.ai | Owners, admins, and members: co-worker chat, files, Brain, integrations, security, oversight |
| MCP gateway | https://mcp.cortexhub.ai/v1/mcp | External AI clients and SDKs |
| Platform API | https://api.cortexhub.ai | Partner backends and automation |
| Documentation | https://docs.cortexhub.ai | Integrators and admins |
The mobile app uses the same conversations, Brain, and governed actions as the web app, and is the most common place to handle approvals on the go.
Who CortexHub is for
- Admins who need to create organization-owned agents, install Slack bots, and control their tools.
- Members who want a personal governed co-worker with a Brain, scheduled routines, and connected business tools.
- End users who want governed tools in their existing AI client without managing credentials in chat.
- Owners and admins who need least-privilege access, approvals, data-handling controls, co-worker oversight, and audit for all AI tool use.
- Enterprises rolling out AI to the whole workforce without shadow-IT risk.
- Developers and partners who embed the same governance in their own products through the SDK and Platform API.