Agent Loop docs

Configuration

Use Settings to validate a mapping before constructing a client. This example creates an OpenAI client with a placeholder credential; constructing and closing it makes no model request:

from agentloop import Settings, build_client

settings = Settings.from_env({
    "LLM_PROVIDER": "openai",
    "OPENAI_API_KEY": "example-key",
    "OPENAI_MODEL": "example-model",
    "LLM_MAX_ATTEMPTS": "2",
})
client = build_client(settings)
try:
    print(client.name, client.model)
    print(f"max attempts: {settings.max_attempts}")
finally:
    client.close()

Output:

openai example-model
max attempts: 2

The application supplies the mapping. Pass os.environ when you want process environment variables; Settings.from_env() does not read them on its own. build_client() applies the configured retry policy, and the caller closes the client it created.

Environment variables

Variable	Default	Meaning
`LLM_PROVIDER`	`openrouter`	`anthropic`, `openai`, `openrouter`, or `modal`
`LLM_MAX_ATTEMPTS`	`3`	Total attempts, including the first request
`LLM_TIMEOUT_SECONDS`	provider default	Read timeout for one model call (Anthropic: the SDK's overall timeout)
`ANTHROPIC_API_KEY`	unset	Anthropic credential
`ANTHROPIC_MODEL`	`claude-opus-5`	Anthropic model
`OPENAI_API_KEY`	unset	OpenAI credential
`OPENAI_MODEL`	`gpt-5.1`	OpenAI model
`OPENAI_API`	`responses`	`responses` or `completions`
`OPENROUTER_API_KEY`	unset	OpenRouter credential
`OPENROUTER_MODEL`	`openrouter/free`	OpenRouter model or router
`MODAL_ENDPOINT_URL`	unset	Base URL of the deployed endpoint
`MODAL_ENDPOINT_MODEL`	`Qwen/Qwen3.6-35B-A3B-FP8`	Model exposed by Modal/vLLM
`MODAL_PROXY_TOKEN_ID`	unset	Optional Modal proxy token ID
`MODAL_PROXY_TOKEN_SECRET`	unset	Optional Modal proxy token secret

You can also construct Settings directly when values come from application configuration or a secrets service:

from pydantic import SecretStr

from agentloop import Settings

settings = Settings(
    provider="openai",
    openai_api_key=SecretStr("example-key"),
    openai_model="example-model",
    max_attempts=2,
)
print(settings.provider, settings.model, settings.max_attempts)

Output:

openai example-model 2

The keys above are placeholders; no request is sent by either example. Replace them with application-supplied credentials before calling a provider. Settings rejects unknown fields, invalid provider names, and nonpositive attempt or timeout values.

See the config API for every field.