@agentproto/adapter-claude-sdk

First-party agentproto adapter that runs Claude Code's agent harness as a library — the Claude Agent SDK's headless query() async generator — behind an AIP-44 ACP server. The daemon spawns it like any other agent-CLI arm; a user can launch it standalone via agentproto-claude-sdk acp.

I/O stays 100% Anthropic-native: the SDK's message stream is relayed to ACP session/updates with no translation of the model I/O.

Why wrap the SDK directly?

The @agentproto/adapter-claude-code adapter wraps the third-party @agentclientprotocol/claude-agent-acp bridge. Driving the SDK directly instead buys us:

• Clean model pinning — options.model, no spawn-arg rejection (issue #186).
• Native usage telemetry — a usage_update per turn (tokens + cost).
• Custom base URL — base_url → ANTHROPIC_BASE_URL, so the same Anthropic-native harness can front real Anthropic, Bedrock/Vertex/Azure, or an Anthropic-compatible gateway (LiteLLM / claude-code-router). Gateway-side translation is out of scope.

Usage

# Spawned by the daemon as the `claude-sdk` arm, or standalone:
agentproto-claude-sdk acp [--model claude-opus-4-8] \
  [--base-url https://gateway.example/v1] [--auth-token <token>] [--thinking]

Auth is read from the spawn env: ANTHROPIC_API_KEY / ANTHROPIC_AUTH_TOKEN, or CLAUDE_CODE_USE_BEDROCK / CLAUDE_CODE_USE_VERTEX / CLAUDE_CODE_USE_FOUNDRY.

Options (AIP-45 manifest)

| id | type | applied as | | ------------ | ------- | ------------------------------------------------------- | | model | string | --model <id> → SDK options.model | | base_url | string | ANTHROPIC_BASE_URL in the child env (see below) | | auth_token | string | ANTHROPIC_AUTH_TOKEN in the child env (Bearer auth) | | thinking | boolean | --thinking → SDK options.thinking = { type: enabled } |

Injected MCP servers (session/new.mcpServers) forward to SDK options.mcpServers, so the daemon can mount a scoped toolset like any other arm.

Anthropic-compatible gateways

base_url + auth_token point one spawn at an Anthropic-compatible gateway with a per-spawn Bearer key (the ambient ANTHROPIC_API_KEY is for real Anthropic). auth_token becomes ANTHROPIC_AUTH_TOKEN, which the SDK sends as Authorization: Bearer <token> — accepted by Moonshot, OpenRouter, and DeepSeek. The token value is never logged.

When base_url is set the adapter enters gateway mode: it pins every model tier the harness might internally request — ANTHROPIC_MODEL, ANTHROPIC_DEFAULT_OPUS_MODEL, ANTHROPIC_DEFAULT_SONNET_MODEL, ANTHROPIC_DEFAULT_HAIKU_MODEL, ANTHROPIC_SMALL_FAST_MODEL — to the resolved model. Without this, a single-model gateway (e.g. Moonshot serving only kimi-k2.7-code) rejects the harness's background claude-haiku-* requests. Native Anthropic (no base_url) leaves tier routing untouched.

Verified live: Moonshot https://api.moonshot.ai/anthropic + model=kimi-k2.7-code, OpenRouter https://openrouter.ai/api, and Requesty https://router.requesty.ai all return valid Anthropic Messages format under Authorization: Bearer.

Note the base URLs carry no /v1: the SDK appends /v1/messages itself, so a /v1 in the preset yields …/v1/v1/messages → 404. Verify a gateway by POSTing to <base_url>/v1/messages — NOT by curling the endpoint you think it serves. An earlier "verified live" note recorded OpenRouter as …/api/v1 on the strength of a direct curl to /api/v1/messages, which left every openrouter-mode spawn 404ing behind the misleading gateway error "model may not exist or you may not have access to it".

Gateway presets (modes)

So you don't hand-type the base URL each spawn, four modes pre-wire the gateway endpoint (base_url/auth_token/thinking still work manually for anything else):

| mode | pre-wires | you supply | | ----------- | --------------------------------------------------------------- | ----------------------------------- | | default | nothing — native Anthropic | ANTHROPIC_API_KEY | | moonshot | ANTHROPIC_BASE_URL + model=kimi-k2.7-code + --thinking | auth_token (Moonshot key) | | openrouter| ANTHROPIC_BASE_URL | model (slug) + auth_token (key) | | requesty | ANTHROPIC_BASE_URL | model (slug) + auth_token (key) | | deepseek | ANTHROPIC_BASE_URL + model=deepseek-v4-pro | auth_token (DeepSeek key) |

mode: moonshot is a one-pick Kimi run; override model for another Moonshot model. mode: openrouter still needs a model (e.g. z-ai/glm-5.2, deepseek/deepseek-v4-pro, moonshotai/kimi-k2). mode: deepseek defaults to deepseek-v4-pro; override model for deepseek-v4-flash. The auth_token is the gateway key — the ambient ANTHROPIC_API_KEY stays for real Anthropic. DeepSeek isn't thinking-gated, so unlike moonshot no --thinking is forced.

Extended thinking

Some gateway models are thinking-gated — kimi-k2.7-code rejects any request that omits thinking (invalid thinking: only type=enabled is allowed for this model). The SDK's query() exposes Options.thinking?: ThinkingConfig ({ type: 'adaptive' } | { type: 'enabled', budgetTokens? } | { type: 'disabled' }); the thinking boolean option pass-through sets options.thinking = { type: 'enabled' }. Off by default so native Claude models keep their own (adaptive) thinking behaviour.

See claude-sdk.ACP.md for the full wire profile.