Sophon CLINEW

The Sophon CLI (sophon) is a cross-platform terminal client for the Sophon AI personal assistant. It talks to the same Gateway as the Dashboard, so your sessions, agents, models, and memory are identical on both surfaces — the CLI is a peer client, not a thin alternative.

Reach for it when you want to script, automate, or stay inside a terminal multiplexer instead of a browser. The CLI shares ~/.sophon/ with the Gateway, so anything you do here shows up in the Dashboard and vice versa.

Installation

The CLI ships on NuGet as the package sophon-cli and installs as a .NET global tool. It runs on Windows, macOS, and Linux and requires the .NET 10 runtime.

dotnet tool install -g sophon-cli

Once installed, the command is sophon. To upgrade or remove it later:

dotnet tool update -g sophon-cli      # upgrade to the latest version
dotnet tool uninstall -g sophon-cli   # remove the tool

Verify the install and see the full command tree:

sophon --version
sophon --help

Two ways to use it

There are two modes of operation:

• Interactive chat REPL. Run sophon with no arguments (or sophon chat) to drop into a streaming terminal chat — token-by-token responses, live plan trees, tool-progress lines, inline approval cards, slash commands, and @-mentions. See Interactive Chat.
• One-shot commands. Run sophon <group> <command> to drive a single surface non-interactively, for example sophon models list or sophon sessions list. Every Dashboard surface — agents, models, approvals, tasks, channels, skills, workflows, cron, documents, memory, and more — is exposed as a subcommand. See the Command Reference.

sophon                 # interactive chat REPL
sophon chat            # same as above
sophon models list     # one-shot command
sophon <group> --help  # discover any group's subcommands

First run

Run the setup wizard the first time you install:

sophon onboard

The wizard checks that the Gateway is reachable, walks you through configuring your first LLM provider (Anthropic, OpenAI, or a local Ollama endpoint), tests the connection, and offers an optional "About You" personalization profile. Provider keys are written encrypted to ~/.sophon/config/models.json.

For headless or CI setups, sophon onboard --non-interactive reads provider settings from environment variables (SOPHON_PROVIDER_TYPE, SOPHON_API_KEY, SOPHON_ENDPOINT), and --skip-profile skips the personalization step.

Quick start

A minimal path from zero to chatting:

dotnet tool install -g sophon-cli   # 1. install
sophon onboard                      # 2. configure a provider and test it
sophon                              # 3. start chatting

After onboarding, running sophon with no arguments always drops straight into the chat REPL. Add your first provider key during the wizard, then iterate from there — see Configuration for the broader setup picture.

In this section

• Interactive Chat — the chat REPL, key bindings, slash commands, and @-mentions.
• Command Reference — every sophon <group> <command>, plus global options.
• Configuration & Auth — the cli.json config file, environment overrides, and login for remote gateways.
• Themes — the four built-in palettes and the automatic no-color mode.
• Scripting & Automation — JSON/CSV output, piping, and non-interactive workflows.