config

Examples

These programs exercise agent-sdk-go (github.com/agenticenv/agent-sdk-go). Agents run as Temporal workflows, so a running Temporal service is mandatory for every example below.

Prerequisite: These examples use agent-sdk-go on the Temporal runtime. A running Temporal server is required before you run them. See Temporal setup for Docker, Temporal CLI, ports, Cloud, and self-hosted options.

Default connection

The examples use TEMPORAL_HOST, TEMPORAL_PORT, and TEMPORAL_NAMESPACE from .env (default: localhost, 7233, default). Adjust if your Temporal runs elsewhere.

Examples overview

Setup

cp env.sample .env
# Edit .env: set LLM_APIKEY, LLM_MODEL (see LLM_PROVIDER: openai, anthropic, or gemini)

Run examples

Minimal agent (no tools)

go run ./simple_agent "Hello, what can you do?"

Agent with caller-owned Temporal client

Uses WithTemporalClient and WithTaskQueue. The example creates the Temporal client, passes it to the agent, and closes it when done. Use this pattern for TLS, Temporal Cloud API keys, or other connection options.

go run ./agent_with_temporal_client "Hello, what can you do?"

Agent with conversation (multi-turn)

Uses in-memory conversation. Run interactive mode (no args) for multi-turn in one process—history is shared across turns. With args, runs a single turn (useful for testing).

# Interactive: type prompts, get responses; history shared. Type 'exit' to end.
go run ./agent_with_conversation

# Single turn (new process each run; no shared history)
go run ./agent_with_conversation "Hello, remember I'm Alice"

Agent with tools

go run ./agent_with_tools "What's the weather in Tokyo?"

Streaming (partial content as tokens arrive)

go run ./agent_with_stream "What's the current time and what's 17 * 23?"

Structured JSON response (WithResponseFormat)

Uses agent.WithResponseFormat with interfaces.ResponseFormatJSON, Name, and interfaces.JSONSchema. No tools—keeps the run in structured-output mode. Prints validated, indented JSON.

go run ./agent_with_json_response
go run ./agent_with_json_response "What is the capital of Japan?"

Reasoning / thinking (WithLLMSampling + LLMReasoning)

Sets WithLLMSampling with Reasoning: &interfaces.LLMReasoning{Enabled, Effort, BudgetTokens} and uses Stream so you can see thinking_delta events when the provider emits them (e.g. Anthropic extended thinking). Pick a model that supports reasoning/thinking for your LLM_PROVIDER.

go run ./agent_with_reasoning
go run ./agent_with_reasoning "Why is the sky blue? One short paragraph."

Streaming + conversation (event handling pattern)

Interactive multi-turn with Stream. Uses AgentEventTypeTextMessageContent (deltas) and AgentEventTypeRunFinished (final body) so the same answer is not printed twice.

go run ./agent_with_stream_conversation
go run ./agent_with_stream_conversation "What is 5 * 8?"

Sub-agents (main agent + specialist)

Two agents in one process: main agent with a math specialist registered via WithSubAgents. Requires workers on both task queues (each NewAgent starts its own embedded worker). Main agent uses default tool approval (RequireAll): delegating to the specialist prompts on stdin (y / n). Specialist uses AutoToolApprovalPolicy so calculator does not prompt. Same stdin pattern as agent_with_tools_approval.

go run ./agent_with_subagents "What is 987 times 654?"

Tools + approval, custom tools, multiple agents, worker split

go run ./agent_with_tools_approval "What is 15 + 27?"
go run ./agent_with_run_async "What is 15 + 27?"
go run ./agent_with_custom_tools "Reverse 'hello world'"
go run ./agent_with_tool_authorizer "Get the protected note for roadmap."
go run ./multiple_agents "What is 7 times 8?"
go run ./multiple_agents concurrent "What is 7 times 8?"

# Agent and worker in separate processes: two terminals — worker in terminal 1,
# agent in terminal 2 (after the worker is up).
go run ./agent_with_worker/worker    # terminal 1: worker
go run ./agent_with_worker/agent "Hello from remote agent!"   # terminal 2: agent

# durable_agent: same two-terminal flow; streaming REPL. Scenarios:
# durable_agent/README.md
go run ./durable_agent/worker       # terminal 1
go run ./durable_agent/agent "Hello from remote agent!"   # terminal 2

MCP (agent_with_mcp_config, agent_with_mcp_client)

Same MCP_* env (see env.sample); differs only in WithMCPConfig vs mcpclient.NewClient + WithMCPClients.

go run ./agent_with_mcp_config
go run ./agent_with_mcp_config "List tools you can call."
go run ./agent_with_mcp_client
go run ./agent_with_mcp_client "List tools you can call."

Configure transports, test against real MCP servers (streamable HTTP walkthrough, stdio, links): agent_with_mcp_config/README.md.

A2A client (agent_with_a2a_config, agent_with_a2a_client)

Outbound A2A tools — set A2A_URL (and optional A2A_* in env.sample).

go run ./agent_with_a2a_config
go run ./agent_with_a2a_config "What tools do you have available?"
go run ./agent_with_a2a_client
go run ./agent_with_a2a_client "What tools do you have available?"

Run a sample remote agent (e.g. a2a-samples helloworld), curl checks: agent_with_a2a_config/README.md.

A2A server (agent_with_a2a_server)

Inbound JSON-RPC server — A2A_SERVER_*, optional bearer tokens.

go run ./agent_with_a2a_server

curl, a2a CLI, testing with agent_with_a2a_config: agent_with_a2a_server/README.md.

Observability OTLP (agent_with_observability)

Requires a reachable OTLP collector (OTEL_EXPORTER_OTLP_ENDPOINT, typically localhost:4317 for gRPC or localhost:4318 for HTTP — see env.sample). Same Temporal + LLM setup as other examples.

go run ./agent_with_observability/config/
go run ./agent_with_observability/objects/

go run ./agent_with_observability/config/ "Say hello in one sentence"
go run ./agent_with_observability/objects/ "Say hello in one sentence"

Details, env semantics, and collector notes: agent_with_observability/README.md.

Vector retriever (agent_with_retriever)

Requires a running vector store (Weaviate or Postgres with pgvector) plus Temporal and LLM env. Set backend-specific vars in env.sample (WEAVIATE_* or PGVECTOR_DSN).

# Weaviate (run ./agent_with_retriever/weaviate/setup.sh; ./cleanup.sh when done)
go run ./agent_with_retriever/weaviate "What is the return policy?"

# pgvector (run ./agent_with_retriever/pgvector/setup.sh; ./cleanup.sh when done)
go run ./agent_with_retriever/pgvector "What is the return policy?"

RETRIEVER_MODE=prefetch go run ./agent_with_retriever/weaviate "What are the return and shipping rules?"

Setup guides: agent_with_retriever/README.md, weaviate/README.md, pgvector/README.md.

Logging

Examples send conversation (user prompt, assistant response) to stdout and internal logs to stderr. By default only errors are logged.

• See logs while evaluating: Set LOG_LEVEL=info or LOG_LEVEL=debug in .env, or run:

  LOG_LEVEL=debug go run ./simple_agent "Hello, what can you do?"
  

• Save logs to a file: Redirect stderr to a file:

  LOG_LEVEL=info go run ./simple_agent "Hello" 2>debug.log
  

• Suppress logs: Show only conversation output:

  go run ./simple_agent "Hello" 2>/dev/null
  

Env vars