Cortex is the agent platform our engagements run on.

What an agent on it does

• Reads across your systems

  The CRM, the support desk, the issue tracker and the call recordings, in one conversation. A renewal brief that takes a person an evening of tabs is one request.

• Asks before it writes

  It drafts the call summary, the next steps or the CRM update, and shows you what it will change before it changes it. Reading needs no sign-off; writing needs a yes.

• Works with the access you give it

  Each agent is connected to the systems its process needs and nothing else, so what it can touch is a decision you make once and can read back later.

The use cases say what that looks like for a revenue team and a customer success team: the problem, the systems the agent works in and how it is measured.

Four ways to run Cortex

Cortex is model-agnostic. Its agents ask for a kind of model (fast, balanced, smart or vision), never a particular vendor’s, and every call goes through one OpenAI-compatible gateway. Run it on Claude, Gemini, models on your own hardware, or anything else the gateway can reach, and change your mind later without touching an agent.

• Open sourceComingRun the whole platform yourself under the Apache-2.0 license: the agent runtime, the server, the CLI and the model gateway setup.
• In your own cloudEarly accessDeploy from AWS Marketplace or Google Cloud Marketplace into your own account. Your network, your models through Bedrock, Vertex AI or your own keys; your data never leaves your account.
• Hosted by usEarly accessWe run it for you. Connect your systems, pick your models, and leave upgrades, scaling and security to us.
• With our helpAvailable nowWe find the processes worth handing to agents, connect Cortex to your systems with your team, and measure what comes back.