Anthropic

RocketRide ships a native Anthropic node that connects Claude models directly to your pipelines — no adapter layer required. It covers every Claude 4 model including Opus, Sonnet, and Haiku, and automatically enables extended thinking for reasoning-capable models.

Nodes

Node	Provider	What it does
`llm_anthropic`	`llm_anthropic`	Generates text via Claude (Sonnet, Opus, Haiku). Streams extended thinking on reasoning-capable models.

Authentication

The node accepts an apikey config field. Use an environment variable so the key stays out of your .pipe file:

{ "config": { "apikey": "${ANTHROPIC_API_KEY}" } }

Set the variable before starting the pipeline:

export ANTHROPIC_API_KEY=sk-ant-...
rocketride start --pipeline ./my-pipeline.pipe

The node validates the key and model at pipeline startup. If the key is invalid or the model is inaccessible, the pipeline reports an error before processing any data. Keys must start with sk-ant.

Choosing a model profile

Set the profile field in config:

{ "config": { "profile": "claude-sonnet-4-6", "apikey": "${ANTHROPIC_API_KEY}" } }

Profile	Model ID	Context	Output
`claude-sonnet-4-6` (default)	`claude-sonnet-4-6`	1M tokens	128K tokens
`claude-opus-4-7`	`claude-opus-4-7`	1M tokens	128K tokens
`claude-opus-4-6`	`claude-opus-4-6`	1M tokens	128K tokens
`claude-sonnet-4-5`	`claude-sonnet-4-5`	1M tokens	64K tokens
`claude-opus-4-5`	`claude-opus-4-5`	200K tokens	64K tokens
`claude-haiku-4-5`	`claude-haiku-4-5`	200K tokens	64K tokens
`custom`	(user-specified)	configurable	—

Choosing a tier:

Haiku — fastest and cheapest; good for classification, routing, and short structured outputs where latency matters.
Sonnet — best balance of speed and capability; the right default for most RAG, summarization, and Q&A pipelines.
Opus — highest capability; use for complex reasoning, long-document analysis, and tasks that benefit from extended thinking.

Extended thinking

When the selected model supports extended thinking (Opus 4.6+, Sonnet 4.6+), the node enables it automatically. The model's reasoning trace is streamed over the thinking SSE lane in parallel with the answers lane. You can wire a downstream node to the thinking lane to capture or display the chain of thought, or ignore it — the answers lane delivers the final response either way.

No extra config is required to enable thinking; it activates based on the selected profile.

Minimal pipeline: chat

{
  "nodes": [
    { "id": "source_1", "provider": "webhook" },
    {
      "id": "llm_1",
      "provider": "llm_anthropic",
      "config": { "profile": "claude-sonnet-4-6", "apikey": "${ANTHROPIC_API_KEY}" },
      "input": [{ "lane": "questions", "from": "source_1" }]
    },
    {
      "id": "target_1",
      "provider": "response",
      "input": [{ "lane": "answers", "from": "llm_1" }]
    }
  ]
}

See the Webhook Pipeline example for a walkthrough of this pattern.

Minimal pipeline: RAG with Anthropic

{
  "nodes": [
    { "id": "source_1", "provider": "webhook" },
    {
      "id": "embed_1",
      "provider": "embedding_openai",
      "config": { "profile": "text-embedding-3-small", "apikey": "${OPENAI_API_KEY}" },
      "input": [{ "lane": "text", "from": "source_1" }]
    },
    {
      "id": "store_1",
      "provider": "qdrant",
      "config": { "profile": "self-hosted", "serverName": "localhost", "collection": "docs" },
      "input": [
        { "lane": "documents", "from": "embed_1" },
        { "lane": "questions", "from": "source_1" }
      ]
    },
    {
      "id": "llm_1",
      "provider": "llm_anthropic",
      "config": { "profile": "claude-sonnet-4-6", "apikey": "${ANTHROPIC_API_KEY}" },
      "input": [{ "lane": "questions", "from": "store_1" }]
    },
    {
      "id": "target_1",
      "provider": "response",
      "input": [{ "lane": "answers", "from": "llm_1" }]
    }
  ]
}

Swapping providers

Because llm_anthropic and every other LLM node share the same lane contract (questions → answers), you can swap providers by changing the provider and config fields — the rest of the pipeline is unchanged.

Nodes​

Authentication​

Choosing a model profile​

Extended thinking​

Minimal pipeline: chat​

Minimal pipeline: RAG with Anthropic​

Swapping providers​

Related​