opencode-claude-auth

Self-contained Anthropic auth provider for OpenCode using your Claude Code credentials — no separate login or API key needed.

How it works

The plugin registers its own auth provider with a custom fetch handler that intercepts all Anthropic API requests. It reads OAuth tokens from the macOS Keychain (or ~/.claude/.credentials.json on other platforms), caches them in memory with a 30-second TTL, and handles the full request lifecycle — no builtin Anthropic auth plugin required. On macOS, multiple Claude Code accounts are detected automatically and can be switched via opencode auth login.

It also syncs credentials to OpenCode's auth.json as a fallback (on Windows, it writes to both %USERPROFILE%\.local\share\opencode\auth.json and %LOCALAPPDATA%\opencode\auth.json to cover all installation methods). If a token is near expiry, it refreshes directly via Anthropic's OAuth endpoint (zero LLM tokens consumed), falling back to the Claude CLI if the direct refresh fails. Background re-sync runs every 5 minutes.

Prerequisites

• Claude Code installed and authenticated (run claude at least once)
• OpenCode installed

macOS is preferred (uses Keychain). Linux and Windows work via the credentials file fallback.

Installation

For Humans

Option A: Let an LLM do it

Paste this into any LLM agent (Claude Code, OpenCode, Cursor, etc.):

Install the opencode-claude-auth plugin and configure it by following: https://raw.githubusercontent.com/griffinmartin/opencode-claude-auth/main/installation.md

Option B: Manual setup

1. Add the plugin to ~/.config/opencode/opencode.json:

   {
     "plugin": ["opencode-claude-auth@latest"]
   }
   

   The @latest tag ensures OpenCode always pulls the newest version on startup. No manual npm install is needed — OpenCode automatically installs npm plugins using Bun at startup.

2. Use it — just run OpenCode. The plugin handles auth automatically using your Claude Code credentials.

For LLM Agents

See installation.md for step-by-step agent instructions.

Usage

Just run OpenCode. The plugin handles auth automatically — it reads your Claude Code credentials, provides them to the Anthropic API, and refreshes them in the background. If your credentials aren't OAuth-based, the plugin falls through to standard API key auth.

Supported models

15 supported models. Run pnpm run test:models to verify against your account.

Credential sources

The plugin checks these in order:

1. macOS Keychain (all Claude Code-credentials* entries — multiple accounts are detected automatically)
2. ~/.claude/.credentials.json (fallback, works on all platforms)

Multiple accounts (macOS)

If you have multiple Claude Code accounts authenticated on macOS, the plugin detects all of them from the Keychain automatically. Each account is labeled by its subscription tier (Claude Pro, Claude Max, etc.).

To switch accounts:

opencode auth login

Select "Switch Claude Code account" and pick the account you want to use. Your selection is persisted across sessions.

If only one account is found, the switcher is hidden and the plugin uses it directly.

Troubleshooting

Diagnostic logging

If you're hitting auth errors that are hard to reproduce, enable debug logging to capture the full auth flow:

export CLAUDE_AUTH_DEBUG=1

Restart OpenCode and reproduce the issue. The plugin writes structured JSON logs to ~/.local/share/opencode/claude-auth-debug.log. All secrets (tokens, API keys) are automatically redacted — the log file is safe to paste into a GitHub issue.

To write logs to a custom path:

export CLAUDE_AUTH_DEBUG=/tmp/claude-auth-debug.log

Disable when done:

unset CLAUDE_AUTH_DEBUG

Long context (1M)

The context-1m-2025-08-07 beta header is not sent by default. Without it, the API caps context at 200k tokens.

To enable 1M context (requires Claude Max or a plan with extra usage coverage), use either of these methods:

Option A: Config file (recommended — no environment setup needed)

Add enable1mContext to any agent in your opencode.json (project-level or ~/.config/opencode/opencode.json). Setting it in any one agent enables 1M context globally for all supported models — you don't need to set it for each agent:

{
  "plugin": ["opencode-claude-auth@latest"],
  "agent": {
    "build": {
      "enable1mContext": true
    }
  }
}

Option B: Environment variable

export ANTHROPIC_ENABLE_1M_CONTEXT=true

If both are set, the environment variable takes priority.

The Claude CLI itself treats 1M context as opt-in (via a [1m] model suffix). Sending the beta without a plan that covers long context charges causes "Extra usage is required for long context requests" errors. Versions before 0.8.0 sent this beta automatically for 4.6+ models, which broke things for Pro users (#64).

If a long context error still occurs (e.g. from a beta flag added via ANTHROPIC_BETA_FLAGS), the plugin retries without the offending flag.

Validating OAuth refresh

To verify the direct OAuth token refresh works with your credentials:

pnpm run validate:oauth           # refresh + write-back (safe, keeps credentials valid)
pnpm run validate:oauth -- --dry-run  # show what would be sent without making the request

This reads your stored credentials, calls Anthropic's OAuth token endpoint, and writes the new tokens back to storage. Refresh tokens rotate on each use, so write-back is enabled by default to keep your stored credentials valid.

Environment variable overrides

All configurable parameters can be overridden via environment variables. If Anthropic changes something before we publish an update, set an env var and keep working: