Anthropic Chat Model

Use Anthropic's Claude models as the LLM in your AI agent workflows

Written By pvdyck

Last updated About 3 hours ago

Anthropic Chat Model

Connects to Anthropic's Claude API to provide LLM capabilities for Agent and Chain LLM workflows.

Authentication

Requires an Anthropic API key configured via Secure Vault. Token usage counts toward your Anthropic API billing.

Parameters

ParameterDescription
ModelThe Claude model to use (see table below).
TemperatureControls randomness (0 = deterministic, 1 = creative). Default: 0.7.
Max TokensMaximum number of tokens in the response. Model-dependent limit.
Top PNucleus sampling threshold (0-1). Alternative to temperature for controlling randomness.

Available Models

ModelBest For
Claude 4 SonnetBest balance of speed, cost, and capability
Claude 4 OpusMost capable, complex reasoning tasks
Claude 3.5 HaikuFastest and cheapest, simple tasks
Claude 3.5 SonnetStrong all-rounder

Model availability depends on your Anthropic API plan.

Options

OptionDescription
System PromptSet persistent instructions for the model's behavior.
Top KLimits token selection to the K most likely next tokens. Lower values make output more focused; higher values allow more variety.
Sampling TemperatureFine-grained temperature control (alternative to the main Temperature parameter). Higher values increase diversity but elevate hallucination risk.

Connects To

Parent NodeDescription
AgentPowers the agent's reasoning and tool-calling decisions.
Chain LLMGenerates text responses from prompts.
Information ExtractorExtracts structured data from text.
Text ClassifierClassifies text into categories.
Sentiment AnalysisAnalyzes text sentiment.

Limitations

  • Must be connected to a parent node — cannot run standalone.
  • No streaming support in the Worker environment.
  • Token limits vary by model (check Anthropic docs for current limits).

Tips

  • Use Claude 4 Sonnet for most workflows — it offers the best cost/performance ratio.
  • Set Temperature to 0 for deterministic, reproducible outputs (e.g., data extraction, classification).
  • Set Temperature to 0.7-1.0 for creative tasks (e.g., content generation).
  • Consider using OpenRouter Chat Model instead for access to Claude plus other providers through a single credential.

Related