Use with AI agents

Install the CLI, then give your coding agent this instruction:

Read https://docs.appaigateway.com/agents.md. Use agw to configure my gateway.
Read before changing resources, use JSON output, and validate app configuration.
Give me browser handoff URLs for provider secrets and account claim. I will open
those outside your browser session. Never print credentials or run inference
unless I explicitly ask.

The first provider or app creation can initialize a new hosted account. Existing accounts use agw account login; your own deployment can be installed with agw deployment setup. Help, lists and status do not create accounts.

For a provider credential, the agent can run:

agw provider add --type openai --browser --no-open --json

Open the returned URL in your own browser, check the provider and destination, and enter the credential. The agent polls the operation and receives only its nonsecret result. Hidden terminal prompts prevent echo, but do not isolate a credential from an agent controlling that terminal.

Account login and claim open in your own browser too. The page names both the account being claimed and the person signed in to approve it; compare them with the terminal before approving. The person claiming creates their sign-in on that page; if the browser is already signed in as someone with an account, the page asks them to sign out first. Claiming keeps the CLI connected, so it can go on configuring resources and reading usage; human-only administration still needs you. Revoke its key from the console whenever you want that access to end.

Application keys go to private files, never standard output. The CLI preserves retry proofs and credentials in protected local state. Keep that state after an interrupted request so retries recover the same operation or resource.

Machine-readable instructions

• /agents.md: the complete agent manual as plain markdown.
• /llms.txt: an index of the documentation.
• /llms-full.txt: the complete documentation in one file.
• /openapi.json: the API contract.
• Each guide as markdown by appending .md to its path.

These are static files. Reading a guide does not install a skill or change your project instructions. You may link the manual from your own AGENTS.md if useful.