Guides

Create agents from templates, GitHub repos, or from scratch

Run agents on a pluggable runtime — Claude Code (default), OpenAI Codex, or Gemini CLI

Start, stop, rename, delete agents and monitor health

The stateless per-turn chat on the agent page — real-time status, playbook autocomplete, attachments, and Stop

A surface the agent keeps current: blocks, the design kit, starter layouts, share links, and PDF export

Ephemeral key-based SSH access into an agent container; the WebSocket terminal for API clients

What a resuming conversation preserves, how auto-compact works, the 50-turn agentic-loop cap, and how the stateless Chat tab differs from the Workspace

File browser, virtual filesystem, and shared folders

Declare runtime data paths, keep data out of git, and export/import an agent's data to move it between Trinity instances

Log viewing, telemetry, and Vector aggregation

Autonomy levels, read-only mode, resources, and timeout settings

Claude, GitHub, email (Resend) and Gemini keys, set from the browser

Inject, hot-reload, and back up agent credentials, plus the Credential Vault

OAuth2 helper API for Google, Slack, GitHub, and Notion — authorization URLs, not a completed exchange

Shared Claude subscriptions, auto-assign, auto-switch, headroom and weekly-limit alerts

Connect Claude Code and other MCP clients to Trinity

A2A 0.3.0 in both directions: inbound tasking, and outbound calls to external agents

Command-line interface for managing agents from your terminal

Cron schedules, execution queue, and misfire handling

Background tasks with activity tracking and result injection

Parallel task dispatch with aggregated results

Bounded sequential task repetition with templates, stop signals, and deadline, budget, and no-progress guardrails

One-shot, durable, agent-initiated deferred follow-ups — an agent schedules a future re-invocation of itself that survives restarts

Multi-source skills library, assignment, injection, auto-sync, and playbooks

Human-in-the-loop approval gates for agent actions — answered from the Operations queue, the mobile admin, or the Workspace user an ask is addressed to

Keep agents in sync with GitHub repositories

Connect agents to Slack with DMs, mentions, and multi-agent routing

x402 payment protocol for agent monetization

Share agents with other users and set access levels

Cross-channel access control with email verification and approval workflows

Public chat URLs, email verification, and session memory

Signed-in chat app — chat tabs and a pinned Main, the composer, the rail, rooms, voice, and file upload

Tags, filtering, and system views

Mobile PWA at /m — agent status, the operator queue, autonomy toggles, and fleet controls

Voice chat, VoIP telephony, image generation, agent avatars, and dynamic dashboards

Real-time voice calls inside a Workspace chat — the orb, the agent's canvas, the transcript in the thread

Agents place outbound phone calls via Twilio + Gemini Live

Agents speak channel replies as ElevenLabs voice notes on Telegram, Slack, and WhatsApp — fail-soft to text

Gemini two-step image pipeline for agents

AI-generated avatars with emotion variants

Custom agent dashboards via YAML configuration

Deterministic safety enforcement for autonomous agent execution

Configure a GitHub Personal Access Token for repository access

Connect agents to Telegram bots with DMs, groups, and @mentions

Connect agents to WhatsApp via Twilio — sandbox and production senders, email verification, and cross-channel access policies

Curated Claude Code plugins covering the full agent lifecycle — scaffold, develop, deploy, and iterate with 5 official plugins