Skip to content

Resources

Documentation

A compact path through the real Gunmetal workflow: install, setup, local API compatibility, client config, and recovery.

Step 01

Install

Install the CLI globally. The npm wrapper installs the native runtime into your local Gunmetal home.

npm i -g @dhruv2mars/gunmetal
  • Public entrypoint for the CLI.
  • Use `gunmetal setup` after install.
Step 02

Set up one provider

Run the guided setup. It connects one provider, checks auth, syncs models, creates one local key, and prints the first test command.

gunmetal setup
  • Browser-session providers open auth.
  • API-key providers store upstream keys locally.
Step 03

Run the local API

Use the Web UI when you want the browser operator surface. Use start when you only need the OpenAI-compatible API.

gunmetal web
# http://127.0.0.1:4684/webui
# http://127.0.0.1:4684/v1
  • Web UI and API share the same daemon.
  • Request history stays local.
Step 04

Point one app at Gunmetal

Any client that accepts a custom OpenAI-compatible base URL, API key, and arbitrary model id can talk to Gunmetal.

OPENAI_BASE_URL=http://127.0.0.1:4684/v1
OPENAI_API_KEY=gm_your_local_key
MODEL=provider/model
  • Use a Gunmetal key, not an upstream provider key.
  • Models use provider prefixes.

API

Contract

GET

/v1/models

List synced provider models visible to the key.

POST

/v1/chat/completions

OpenAI-compatible chat completion endpoint.

POST

/v1/responses

Responses-style endpoint for compatible providers.