Trinity CLI
Command-line interface for Trinity, the operating system for the AI-native company. Manage agents from your terminal with shell commands — a subset of what the MCP server offers (see Limitations).
From Zero to Deployed AI Agent
Apr 2026
Installation
# PyPI (recommended; Python 3.10+) pip install trinity-cli # Homebrew (macOS/Linux) brew install abilityai/tap/trinity-cli # Verify installation trinity --version
First-Time Setup
Run trinity init to connect to a Trinity instance:
trinity init
You'll be prompted for:
trinity.example.com)The CLI stores credentials in ~/.trinity/config.json (mode 0600) and auto-provisions an MCP API key. Options: --profile <name> saves the instance under a named profile, and --admin logs in with the admin password instead of an email code.
Multi-Instance Profiles
Manage multiple Trinity instances (local, staging, production) with named profiles:
# List all profiles trinity profile list # Switch active profile trinity profile use production # Remove a profile trinity profile remove staging
Profile resolution priority:
TRINITY_URL / TRINITY_API_KEY environment variables--profile <name> flagTRINITY_PROFILE environment variablecurrent_profile in config fileReturning Users
Re-authenticate with an existing profile:
trinity login # email code; --admin for the admin password trinity login --instance https://trinity.example.com --profile staging # add or repoint a profile trinity status # which profile and instance you are on trinity logout # clear the current profile's stored credentials
Agent Management
# List all agents trinity agents list # Get agent details trinity agents get my-agent # Create from GitHub template (created as an agent: a working branch with # auto-sync when your own token can push to your own repo, else pull-only) trinity agents create my-agent --template github:user/repo # Start/stop agents trinity agents start my-agent trinity agents stop my-agent # Rename an agent trinity agents rename old-name new-name # Delete an agent trinity agents delete my-agent
Deploy
Deploy a local agent directory to Trinity:
# Deploy current directory trinity deploy . # Deploy with custom name trinity deploy . --name my-agent # Deploy from GitHub trinity deploy --repo user/repo
A directory deploy packages the folder (in a git repo, tracked files plus untracked files .gitignore does not exclude; always leaving out .git, node_modules, virtualenvs, and .env, .env.local, .env.production), adds an integrity manifest, and uploads it — the archive must stay under 50 MB. The agent name comes from --name, then template.yaml's name, then the directory name. The first deploy writes .trinity-remote.yaml, so later deploys from the same directory update the same agent; if that file points at a different instance than your current profile, the CLI asks before deploying. An agent deployed from local files has no repository behind it and no auto-sync.
--repo creates the agent from github:user/repo as a deployment — pull-only, tracking the branch, never pushing back. To give an agent a working branch on your own repo, use trinity agents create … --template github:user/repo instead. A private repo needs a GitHub token on the instance that can read it.
Chat and Logs
# Send a message trinity chat my-agent "Hello, what can you do?" # View chat history trinity history my-agent # View container logs (default: last 50 lines) trinity logs my-agent --tail 200
Health Monitoring
# Fleet-wide health status trinity health fleet # Single agent health trinity health agent my-agent
Skills and Schedules
# List available skills trinity skills list # Get skill details trinity skills get skill-name # List agent schedules trinity schedules list my-agent # Trigger a schedule manually trinity schedules trigger my-agent schedule-id
trinity skills list prints each skill's name, description, category and a deprecated column: yes, → <successor> when the library names a replacement, or blank for a live skill. --format json returns the listing as the API sends it, with deprecated and superseded_by fields. Text from a library author is printed as-is, so a description containing brackets no longer breaks the table.
Tags
# List all tags trinity tags list # Get tags for an agent trinity tags get my-agent
Output Formats
# Table output (default) trinity agents list # JSON output (for scripting) trinity agents list --format json
--format is an option on each listing or read command, not a global flag.
Environment Variables
| Variable | Description |
|---|---|
| TRINITY_URL | Override instance URL |
| TRINITY_API_KEY | Override authentication token |
| TRINITY_PROFILE | Set active profile |