Skip to main content
Kontext connects your team’s agents to one workspace. It checks each tool call against your policies before the call runs. The dashboard shows what happened and the activity’s risk and cost. On each endpoint, the Kontext daemon receives tool calls from agent hooks. The daemon makes the policy decision locally and records the decision and outcome. The workspace receives those records and holds your team’s policies.

The endpoint daemon makes the decision. The workspace receives activity records and supplies policies.

What you find in the dashboard

How policy works

Kontext evaluates policies on the endpoint inside the agent’s pre-tool hook. The agent waits for the decision. The workspace takes no part in that step. A new policy starts in the Observing list. You can see how many calls it would block before it blocks anything. When the matches look right, enforce the policy. In Enforce mode, a Deny stops the call before it runs. Kontext also reports risk and cost for the same activity. Neither changes a decision. How decisions work explains the tool call sequence and failure behavior.

Two ways to deploy

  • To set up one endpoint, open Get started and create a personal API key. Run brew install kontext-security/tap/kontext, then kontext setup. See the quickstart.
  • To set up many endpoints, download the signed package and scripts from Deployments. Push them with your MDM. See deploying Kontext.

Where to go next

Core concepts

Learn how Kontext connects agents, tool calls, and policies to risk and cost.

Supported agents

See what Claude Code, Codex, and Claude Cowork report to Kontext.

Policies

Review what Block GitHub force pushes would block in the Observing list before you enforce it.

Management API

Use the API to read and change policies or manage API keys.