Skip to main content

Zed with Verdictan

Zed supports remote MCP servers and custom OpenAI-compatible providers. Configure one path independently. You can also use the two paths together.

Add Verdictan MCP tools

Open Settings → AI → MCP Servers. Select Add Remote Server. Specify these values:

  • URL: https://<published-hostname>/mcp
  • Header: Authorization: Bearer <client-api-token>

Use Zed's protected settings flow for the credential. Do not commit a raw token. After Zed loads its tools, verify that the server indicator is active.

See Zed's MCP documentation for the UI and tool-permission controls.

Route Zed model requests

First start the local gateway as shown in Gateway Setup for IDEs. Then open Zed's Agent Settings. Add an OpenAI-compatible provider. Specify these values:

FieldValue
Provider nameverdictan
API URLhttp://127.0.0.1:41002/v1
Model IDA model returned by GET /v1/models
API keyA Verdictan client API token
Context windowThe value documented by the selected upstream model

Zed stores provider keys from its settings UI in the system keychain. Select the new provider/model for the Zed feature to test. Zed's edit prediction and external-agent paths have different configuration boundaries. A model selected for Zed Agent does not automatically apply to them.

See Zed's API-access documentation for the OpenAI-compatible provider fields.

Verify

Run verdictan events tail --since 10m --follow. Select the Verdictan model. Send a Zed Agent request. Make sure that the event is available. Test MCP tool discovery independently.

Next steps