The $ marks a terminal prompt; leave it out when typing commands.
Generated from the CLI source in this checkout (package version 0.7.0). Use agentic-api --help or agentic-api <command> --help for your installed version.
Agentic API provides two command-line interfaces (CLIs), both used from a terminal. The PyPI package includes both:
- Python launcher:
agentic-api ...— start the gateway withagentic-api serve, check your installation withagentic-api doctor, or inspect versions withagentic-api version. You can also usepython -m agentic_api .... This page documents the Python launcher. - Rust CLI:
agentic ...— run harnesses withagentic run codexoragentic run claude, or useagentic serveandagentic validate. It has its own commands and flags; runagentic --helpfor its command-line reference.
See Python installation and workflows for package installation, remote and local serving examples, and diagnostics.
agentic-api
usage: agentic-api [-h] [--version] {serve,doctor,version} ...
Python launcher for packaged Agentic API binaries
positional arguments:
{serve,doctor,version}
serve Launch Agentic API in local or remote mode
doctor Report packaged binary and compatibility diagnostics
version Print package and packaged binary versions
options:
-h, --help show this help message and exit
--version show program's version number and exit
agentic-api serve
usage: agentic-api serve [-h] [--model MODEL] [--vllm-base-url VLLM_BASE_URL]
[--host HOST] [--port PORT]
[--startup-timeout-s STARTUP_TIMEOUT_S]
[--shutdown-timeout-s SHUTDOWN_TIMEOUT_S]
[--vllm-port VLLM_PORT]
[--gateway-api-key-env GATEWAY_API_KEY_ENV]
[--vllm-api-key-env VLLM_API_KEY_ENV]
...
Launch the packaged Rust gateway. Choose exactly one of --model (managed local vLLM) or
--vllm-base-url (an existing upstream server).
positional arguments:
vllm_args Extra local vLLM arguments after --
options:
-h, --help show this help message and exit
--model MODEL Model to serve with managed local vLLM; requires a local vLLM
install (default: None)
--vllm-base-url VLLM_BASE_URL
HTTP(S) base URL of an existing upstream, without credentials
(default: None)
--host HOST Gateway listen address (default: 0.0.0.0)
--port PORT Gateway listen port (default: 9000)
--startup-timeout-s STARTUP_TIMEOUT_S
Startup readiness timeout in seconds (default: 600.0)
--shutdown-timeout-s SHUTDOWN_TIMEOUT_S
Grace period in seconds before terminating child processes
forcibly (default: 10.0)
--vllm-port VLLM_PORT
Managed local vLLM listen port (default: 8000)
--gateway-api-key-env GATEWAY_API_KEY_ENV
Environment variable containing the upstream API key used by the
gateway (default: OPENAI_API_KEY)
--vllm-api-key-env VLLM_API_KEY_ENV
Environment variable containing the managed local vLLM API key;
also passed to the gateway (default: AGENTIC_VLLM_API_KEY)
Pass extra vLLM arguments after -- in local mode. The launcher reserves --host, --port,
and --api-key; --uds is incompatible. Ports must be 1-65535. Timeouts must be greater
than 0 and at most 86400 seconds.
agentic-api doctor
usage: agentic-api doctor [-h] [--mode {local,remote}] [--json]
Check the packaged gateway and local vLLM compatibility. Without --mode, report both
modes and use remote health for the exit status. Exit 0 when healthy, 1 otherwise.
options:
-h, --help show this help message and exit
--mode {local,remote}
Select which mode to check (default: None)
--json Emit a machine-readable JSON report (default: False)
agentic-api version
usage: agentic-api version [-h]
Print the Python package, packaged Rust binary, and supported and installed vLLM
versions. Use agentic-api --version to print only the Python package version.
options:
-h, --help show this help message and exit