agentproto provider-preset
agentproto provider-preset list [--json]List the built-in provider gateway presets shipped in
@agentproto/provider-presets (moonshot, openrouter, requesty, opencode-go,
opencode, deepseek, xai, xai-anthropic, llm-endpoint, openai, openai-direct,
mistral, groq, nebius, huggingface, deepinfra), each with its live key-env
status. A preset is static data a claude-code / claude-sdk
agent can front via a mode or the base_url option — there is no install step,
no setup pipeline, and no credentials store.
Requires a running daemon (serve.md or
daemon.md). Discovery reads ~/.agentproto/runtime.json, same
as tunnel list.
Status
Status is answered from the daemon's process.env — where agents actually
spawn — not from your shell. A preset can read ready in the daemon while the
key is unset in the terminal you typed the command into, and vice versa.
| Status | Meaning |
|---|---|
ready | The preset's API-key env var is set (e.g. MOONSHOT_API_KEY). |
available | The env var isn't set. Export it in the daemon's environment to make the preset ready, or pass an auth_token at spawn. |
A key env var can be shared by two presets, in which case both read the same
status off it. opencode-go (the flat OpenCode Go subscription) and opencode
(OpenCode Zen, pay-as-you-go) both declare OPENCODE_API_KEY — opencode's own
convention — so one exported key flips both to ready even though only one
of the two balances is actually funded. They are separate billing rails with
separate secrets; which one a session bills is decided by its auth profile's
endpoint (--access-profile), not by the shared env name.
Subverbs
list
Lists every preset via GET /presets.
| Flag | Default | Description |
|---|---|---|
--json | false | Emit the raw AdapterEntry<PresetInfo>[] array instead of the table. |
The table prints one row per preset: ID, STATUS, SCHEMA, KEY ENV,
DEFAULT MODEL, and BASE URL. SCHEMA is the preset's wire flavor —
anthropic for the gateways claude-code/claude-sdk front directly via
ANTHROPIC_BASE_URL, openai for one reached through the local
OpenAI-compatible proxy. A preset with no default model prints —; pick one
per session with --model.
Examples
# What's shipped, and which keys the daemon can see
agentproto provider-preset list
# Machine-readable — e.g. which presets are ready
agentproto provider-preset list --json | jq -r '.[] | select(.status=="ready") | .slug'ID STATUS SCHEMA KEY ENV DEFAULT MODEL BASE URL
moonshot ready anthropic MOONSHOT_API_KEY kimi-k2.7-code https://api.moonshot.ai/anthropic
openrouter available anthropic OPENROUTER_API_KEY — https://openrouter.ai/api
requesty available anthropic REQUESTY_API_KEY — https://router.requesty.ai
opencode-go available anthropic OPENCODE_API_KEY minimax-m3 https://opencode.ai/zen/go
opencode available anthropic OPENCODE_API_KEY claude-sonnet-4-6 https://opencode.ai/zen
deepseek available anthropic DEEPSEEK_API_KEY deepseek-v4-pro https://api.deepseek.com/anthropic
xai available openai XAI_API_KEY grok-4.5 http://localhost:18090/v1
xai-anthropic available anthropic XAI_API_KEY grok-4.5 https://api.x.ai
llm-endpoint available anthropic LLM_ENDPOINT_ACCESS_TOKENS kimi-k2.7-code http://localhost:18090
openai available openai OPENAI_API_KEY gpt-4.1 http://localhost:18090/v1
openai-direct available openai OPENAI_API_KEY gpt-4.1 https://api.openai.com/v1
mistral available openai MISTRAL_API_KEY mistral-large-latest https://api.mistral.ai/v1
groq available openai GROQ_API_KEY — https://api.groq.com/openai/v1
nebius available openai NEBIUS_API_KEY — https://api.studio.nebius.com/v1
huggingface available openai HF_TOKEN — https://router.huggingface.co/v1
deepinfra available openai DEEPINFRA_API_KEY — https://api.deepinfra.com/v1/openaiExit codes: 2 when no daemon is found (or an unknown subcommand is passed),
1 when the /presets request fails.