Docs · LLMs and developers
Reviewed2026-07-22Version3.12.1Warp terminal
Use Warp's Custom Router with Achiral's memory inference endpoint.
Warp's Custom Router can point to Achiral's memory inference endpoint.
That lets Warp send prompts through your organization's Achiral context instead of using a generic endpoint.
What you need
- an Achiral organization slug, such as
acme; - a Context Access Token with the
inference:chatscope; and - your organization's memory endpoint, such as
https://acme.achiral.ai/v1.
Create and manage these tokens from the organization dashboard.
Setup
- Open your organization dashboard.
- Create a new Context Access Token.
- Make sure the token includes Inference Chat.
- Copy the raw token when the app shows it.
- In Warp, open Settings → AI → Custom Router.
- Set the base URL to
https://{your-slug}.achiral.ai/v1. - Set the authorization header to
Bearer {your-token}. - Save the settings.
What happens next
When Warp sends a prompt to that endpoint, Achiral handles the request through the same memory-aware inference gateway described in Inference gateway.
Token scopes
Only inference:chat is required for the Warp setup on this page.
Other scopes can exist on the token, but they are not required just to use Warp as a memory-aware chat client.
Troubleshooting
403 Forbidden
Check these first:
- the token includes
inference:chat; - the token belongs to the same organization as the URL you used in Warp; and
- the organization is active.
404 Not Found
Check that Warp is pointing at the correct /v1 endpoint for your organization.
Request succeeds but responses are not what you expect
Confirm that:
- you are using the right organization endpoint;
- the organization has the context you expect; and
- the token belongs to the correct organization.