Getting Started

Install Oh My OpenAgent, configure providers, and verify your first OpenCode session.

Prerequisites

Item	Requirement
OpenCode	`opencode` should be installed. The upstream guide expects OpenCode `1.0.150` or newer.
Runtime	The CLI ships standalone binaries for major platforms after install, but `bunx` or `npx` is still the easiest bootstrap path.
Providers	Claude, OpenAI/ChatGPT, Gemini, GitHub Copilot, OpenCode Go/Zen, Kimi for Coding, Z.ai, Vercel AI Gateway, or the subset you actually use.
OS	macOS ARM64/x64, Linux x64/ARM64 including musl/Alpine, and Windows x64 are published.

Install

The upstream package is still published as oh-my-opencode:

bunx oh-my-opencode install

The compatibility binary also exists:

bunx oh-my-openagent install

The installer registers the OpenCode plugin entry as oh-my-openagent when possible, creates an OmO config file, and prints the authentication steps needed for the provider flags you selected.

Non-Interactive Install

Use non-interactive mode when an agent is doing setup for you:

bunx oh-my-opencode install --no-tui \
  --claude=yes \
  --openai=yes \
  --gemini=no \
  --copilot=no

Common provider flags:

Flag	Meaning
`--claude=no\|yes\|max20`	Claude subscription mode. `max20` means Claude Max 20x.
`--openai=no\|yes`	OpenAI / ChatGPT subscription for GPT-backed agents such as Oracle.
`--gemini=no\|yes`	Gemini integration.
`--copilot=no\|yes`	GitHub Copilot fallback provider.
`--opencode-go=no\|yes`	OpenCode Go subscription for GLM, Kimi, MiniMax models.
`--opencode-zen=no\|yes`	`opencode/` model catalog access.
`--zai-coding-plan=no\|yes`	Z.ai Coding Plan, mainly GLM fallbacks.
`--kimi-for-coding=no\|yes`	Kimi for Coding subscription.
`--vercel-ai-gateway=no\|yes`	Vercel AI Gateway provider path.
`--skip-auth`	Skip auth hints after writing config.

Provider selection is agent-specific. OmO does not use one global provider priority; each agent and category has its own fallback chain.

Authenticate Providers

After installation, use OpenCode auth for the providers you enabled:

opencode auth login

For Gemini Antigravity OAuth, upstream currently expects the opencode-antigravity-auth plugin and full model configuration from that plugin's README. For GitHub Copilot, Z.ai, OpenCode Go, OpenCode Zen, and Kimi, the installer chooses model defaults according to the provider flags you passed.

Verify

opencode --version
bunx oh-my-opencode doctor

doctor checks system, config, tools, and model resolution:

Group	Examples
System	OpenCode binary, version, plugin registration, legacy package warnings.
Config	JSONC parsing, schema validation, config file detection.
Tools	AST-Grep, LSP servers, GitHub CLI, MCP servers.
Models	Capability cache, model normalization, agent/category overrides, fallback availability.

Use doctor --verbose when provider/model resolution looks wrong.

First Task

For autonomous work:

ulw fix the failing login flow and verify it in the browser

For precise multi-step work:

@plan refactor the auth module without behavior changes
/start-work

For GPT-native deep work, switch to Hephaestus from the agent selector and give a goal rather than a recipe.

Telemetry

Anonymous active-install telemetry is enabled by default. It sends at most one event per UTC day per machine using a hashed installation identifier; it does not create PostHog person profiles. Disable it with either variable:

OMO_SEND_ANONYMOUS_TELEMETRY=0
OMO_DISABLE_POSTHOG=1