Command-line interface

The Seepient CLI is designed for non-interactive automation, shell scripts, and CI/CD pipelines where a full-screen interactive interface is impractical.

When to use CLI vs TUI vs SDK

• Use the TUI (seepient) when you are actively writing code, reviewing diffs, and collaborating interactively with the agent in a terminal.
• Use the CLI (seepient "prompt") when you want a quick one-shot answer, need to pipe data through standard Unix streams, or want to run automated batch jobs in a shell script.
• Use the SDK (createSeepient) when you are embedding agent behavior into a Node.js web server, desktop app, or serverless worker.

---

One-shot execution

To execute a task and exit immediately upon completion, pass the prompt string as an argument:

seepient "Find all TODO comments in src/ and output a prioritized markdown list"

Seepient initializes the agent loop, executes the necessary tool calls (such as searching files or reading code), prints the formatted response to standard output, and exits with code 0.

Non-interactive automation (-y)

By default, effectful operations (like modifying files or running shell commands) prompt for confirmation. In automated environments where no human is present, use the -y or --no-interactive flag to auto-approve safe actions:

seepient -y "Update all dependencies in package.json to latest compatible versions"

When combined with sandboxed execution, -y runs commands safely inside the macOS Seatbelt or Linux Bubblewrap jail without human intervention.

Overriding providers and models

You can override default configuration directly from the command line:

# Use a specific provider and model
seepient --provider anthropic --model claude-3-7-sonnet "Audit this codebase"

# Route to the high-reasoning complex tier
seepient --tier complex "Resolve the race condition in the worker scheduler"