Pi (experimental)
Configure Pi Coding Agent in Happier (ACP), including install and detection.
Happier can run Pi Coding Agent through the local pi CLI.
Pi is an experimental integration. It works and people use it daily, but its surface is younger than Claude, Codex or OpenCode and is more likely to change. What it supports today is in Agent capabilities.
Install Pi
Happier can install it for you:
happier install provider pi
pi --versionOr install manually:
npm install -g @earendil-works/pi-coding-agent
pi --versionStart Pi with Happier
happier piLocal authentication in Happier
Pi does not currently expose a native interactive login flow through Happier provider settings.
Instead, Happier shows the machine-local auth state for Pi in:
- Settings → AI & Agents → AI backends → Pi
Today that status is based on supported environment-backed credentials on the selected machine.
For the shared user guide, see Provider authentication.
Authentication (connected services)
Pi can reuse connected services profiles when spawned by Happier:
openai-codex(subscription OAuth)openai(API key)claude-subscription(OAuth or setup-token) for Claude subscription authanthropicfor Anthropic API key auth (ANTHROPIC_API_KEY)
Happier materializes compatible connected services into Pi's isolated PI_CODING_AGENT_DIR/auth.json before launch. Claude subscription OAuth profiles are written as Pi OAuth credentials; setup-token and API-key profiles are written as Pi API-key credentials.
If multiple compatible profiles exist, Happier will let you pick one when starting a new session.
Policy note: Claude subscription OAuth credentials are intended for ordinary Claude subscription use in native Anthropic applications. Developers building third-party products should use Anthropic API keys or supported cloud providers unless their policy/legal review permits this Pi integration path.
See: Connected services
Troubleshooting
If Pi is not detected:
- Verify
pi --versionworks. - Restart Happier.
- Set
HAPPIER_PI_PATHif your binary is not inPATH.
Related
- Agent capabilities — what this agent supports, generated from the manifest.
- Choosing an agent
- Model and engine selection
- Agent authentication