Examples
Coding Agent (Anthropic)
A readline-based coding agent that uses Anthropic-specific thinking, caching, beta features, retries, and rate limiting.
examples/coding-agent-anthropic.rs is the most feature-rich coding-agent sample in the crate. It wires up the shared file and shell tools, streams content and reasoning to the terminal, and layers in Anthropic-only controls on top of a normal AgentBuilder.
Run
export ANTHROPIC_API_KEY="sk-ant-..."
cargo run --example coding-agent-anthropicWhat this example actually configures
LlmProvider::Anthropicwith modelclaude-sonnet-4-5- Extended thinking with
ThinkingConfig::enabled(1024) - Prompt caching with a one-hour TTL
- Anthropic beta features for fine-grained tool streaming, context management, interleaved thinking, and 1M context
- Anthropic tool choice, rate limiting, and retry configuration
- Four local tools:
read_file,write_file,bash, andlist_files
Key builder setup
let agent = AgentBuilder::new("claude-advanced")
.provider(LlmProvider::Anthropic)
.model("claude-sonnet-4-5")
.system_prompt(
"You are an advanced coding assistant powered by Claude Sonnet 4.5. \
You have access to file operations, bash commands, and directory listing. \
Use your extended thinking capabilities to reason through complex problems. \
Always explain your reasoning process and provide detailed analysis.",
)
.thinking(appam::llm::anthropic::ThinkingConfig::enabled(1024))
.caching(appam::llm::anthropic::CachingConfig {
enabled: true,
ttl: appam::llm::anthropic::CacheTTL::OneHour,
})
.beta_features(appam::llm::anthropic::BetaFeatures {
fine_grained_tool_streaming: true,
context_management: true,
interleaved_thinking: true,
context_1m: true,
..Default::default()
})
.tool_choice(appam::llm::anthropic::ToolChoiceConfig::Auto {
disable_parallel_tool_use: false,
})
.rate_limiter(appam::llm::anthropic::RateLimiterConfig {
enabled: true,
tokens_per_minute: 1_800_000,
})
.retry(appam::llm::anthropic::RetryConfig {
max_retries: 5,
initial_backoff_ms: 2000,
max_backoff_ms: 60000,
backoff_multiplier: 2.0,
jitter: true,
})
.with_tool(Arc::new(read_file()))
.with_tool(Arc::new(write_file()))
.with_tool(Arc::new(bash()))
.with_tool(Arc::new(list_files()))
.max_tokens(20000)
.build()?;Runtime behavior
- The loop is interactive and exits on
exit,quit, orbye. - The file is standalone; there is no shared
examples/commonhelper. - The loop keeps
Option<Vec<ChatMessage>>, usesagent.initial_messages(input)?on the first turn, appends follow-ups withChatMessage::user(input), and recordssession.messagesafter successful turns. - The response is streamed with
default_run_streaming_with_messages(...)andConsoleConsumer::new(), so content, reasoning, tool calls, and tool results appear in the terminal.