Using Downsizing with BoltAI

Prerequisites

BoltAI installed on macOS or Windows, and a Downsizing account that can make requests (pay-per-use or a connected subscription both work). Grab your Router ID and API key by running ds link boltai in a terminal, or from the Inference API page.

BoltAI manages custom AI providers through its own app storage, not a plain config file on disk — there's nothing for ds to write to directly. Instead, the dsCLI prints the exact values you paste into BoltAI's OpenAI-compatible Server form, and BoltAI stores the connection itself.

1. Get your setup values

Run the following to print the values you'll paste into BoltAI:

ds link boltai

This prints a Chat Completions Endpoint URL, a Model ID, and an Authorizationheader value — everything BoltAI's custom server form asks for. Safe to run more than once; it doesn't write any files.

2. Add an OpenAI-compatible Server in BoltAI

1. Open BoltAI and go to Settings → Models, then click the (+) button and choose OpenAI-compatible Server.
2. Set Friendly Name to something recognizable, e.g. Downsizing.
3. Set Chat Completions Endpoint URL to your Downsizing OpenAI-compatible Chat Completions route:

http://localhost:3001/YOUR_ROUTER_ID/openai/v1/chat/completions

Replace YOUR_ROUTER_ID with the Router ID printed by ds link boltai(or from the dashboard). BoltAI's OpenAI-compatible Server form is documented as expecting a literal Chat Completions endpoint (e.g. ending in /v1/chat/completions), so use the /openai/v1/chat/completionsroute rather than the Anthropic Messages route — that's a requirement of BoltAI's client, not a Downsizing limitation.

4. Set Model ID to the model you want to route through Downsizing, e.g. claude-opus-4-8, and adjust Context Length and the streaming toggle to match your chosen model.
5. Add your Downsizing API key as the provider's API key (or under Custom Headers, depending on your BoltAI version):

Authorization: Bearer dws_your_api_key_here

Replace dws_your_api_key_here with your real Downsizing API key (starts with dws_). Click Save Changes, then set this provider as default if you don't want BoltAI to keep using OpenAI directly.

Verify

Start a new chat in BoltAI, pick the Downsizing custom server you just added from the model picker, and send a test message. If you get a response, the connection is wired up correctly. Then check your savings dashboard to see the request appear in your usage metrics.

Removing the connection

Running ds unlink boltaiupdates Downsizing's own record of linked tools, but since BoltAI stores the custom server in its own app storage, you'll also want to open BoltAI → Settings → Models and remove the Downsizing custom server there.