Struct AnthropicConfig 

pub struct AnthropicConfig {

    pub extended_thinking_enabled: bool,
    pub interleaved_thinking_beta: String,
    pub interleaved_thinking_budget_tokens: u32,
    pub interleaved_thinking_type_enabled: String,
    pub tool_search: ToolSearchConfig,
    pub memory: AnthropicMemoryConfig,
    pub effort: Option<ReasoningEffortLevel>,
    pub task_budget_tokens: Option<u32>,
    pub task_budget_beta: String,
    pub thinking_display: Option<ThinkingDisplayMode>,
    pub count_tokens_enabled: bool,
    pub advisor: AdvisorConfig,
    pub fallbacks: AnthropicFallbacks,
}

Anthropic-specific provider configuration

Fields

extended_thinking_enabled: bool

Enable adaptive or extended thinking for Anthropic models When enabled, Claude uses internal reasoning before responding, providing enhanced reasoning capabilities for complex tasks. Only supported by Claude 4, Claude 4.5, and Claude 3.7 Sonnet models. Claude Opus 4.7 uses adaptive thinking instead of budgeted extended thinking. Note: Extended thinking is now auto-enabled by default (31,999 tokens). Set MAX_THINKING_TOKENS=63999 environment variable for 2x budget on 64K models. See: https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking

interleaved_thinking_beta: String

Beta header for interleaved thinking feature

interleaved_thinking_budget_tokens: u32

Budget tokens for extended thinking (minimum: 1024, default: 31999) On 64K output models (Opus 4.5, Sonnet 4.5, Haiku 4.5): default 31,999, max 63,999 On 32K output models (Opus 4): max 31,999 Claude Opus 4.7 ignores this setting and uses adaptive thinking instead. Use MAX_THINKING_TOKENS environment variable to override.

interleaved_thinking_type_enabled: String

Type value for enabling interleaved thinking

tool_search: ToolSearchConfig

Tool search configuration for dynamic tool discovery (advanced-tool-use beta)

memory: AnthropicMemoryConfig

Native Anthropic memory tool configuration.

effort: Option<ReasoningEffortLevel>

Effort level for adaptive thinking/token usage (low, medium, high, xhigh, max) Controls how many tokens Claude uses when responding, trading off between response thoroughness and token efficiency. Unset by default: each model then uses its own default effort (for example medium on Claude Opus 5.5, high on Claude Opus 5). An explicit agent.reasoning_effort or /effort selection takes precedence; a value the active model does not support falls back to that model’s default.

task_budget_tokens: Option<u32>

Optional Anthropic task budget token total for Claude Fable 5/5.1, Opus 5, and Opus 5.5 (not sent for other models). When set, VT Code sends output_config.task_budget = { type = "tokens", total = N } and the required beta header. Anthropic currently requires a minimum of 20,000 tokens.

task_budget_beta: String

Beta header for Anthropic task budgets.

thinking_display: Option<ThinkingDisplayMode>

Controls how thinking content is returned in API responses.

• “summarized”: Thinking blocks contain summarized text (default on Opus 4.6 and earlier).
• “omitted”: Thinking blocks have an empty thinking field (default on Opus 4.7+).
• “updates”: Only the progress updates between tool calls are returned as text (Claude Sonnet 5.5, Claude Opus 5.5 and Claude Fable 5.x; the VT Code default on Claude Sonnet 5.5 and Claude Opus 5.5). Ignored on models without progress updates.

When set, this overrides the model-specific default. See: https://docs.anthropic.com/en/docs/build-with-claude/extended-thinking#controlling-thinking-display

count_tokens_enabled: bool

Enable token counting via the count_tokens endpoint When enabled, the agent can estimate input token counts before making API calls Useful for proactive management of rate limits and costs

advisor: AdvisorConfig

Claude Advisor server-side tool configuration. The advisor pairs a faster executor model with a higher-intelligence advisor model for strategic guidance mid-generation. Only honored for Anthropic models and providers.

fallbacks: AnthropicFallbacks

Server-side refusal fallbacks (fallbacks request parameter).

• “default”: let Anthropic pick the recommended fallback model when the primary model declines on policy grounds (the VT Code default).
• “off”: never send fallbacks.
• a list of 1-3 { model, max_tokens } entries tried in order.

Only sent to the first-party Claude API (api.anthropic.com) for models whose capability profile supports server-side fallbacks (Claude Opus 5, Opus 5.5, Fable 5 and Fable 5.1); other models and endpoints send nothing.

Trait Implementations

impl Clone for AnthropicConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for AnthropicConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for AnthropicConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for AnthropicConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl JsonSchema for AnthropicConfig

fn schema_name() -> Cow<'static, str>

The name of the generated JSON Schema.

fn schema_id() -> Cow<'static, str>

Returns a string that uniquely identifies the schema produced by this type.

fn json_schema(generator: &mut SchemaGenerator) -> Schema

Generates a JSON Schema for this type.

fn inline_schema() -> bool

Whether JSON Schemas generated for this type should be included directly in parent schemas, rather than being re-used where possible using the $ref keyword.

impl Serialize for AnthropicConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.