OpenCode

LNAI exports unified configuration to OpenCode’s native format, transforming settings as needed.

Output Structure

.opencode/
├── AGENTS.md          # Symlink → ../.ai/AGENTS.md
├── rules/             # Symlink → ../.ai/rules/
├── skills/<name>/     # Symlinks → ../../.ai/skills/<name>/
└── <overrides>        # Symlinks from .ai/.opencode/

opencode.json          # Generated (at project root)

File Mapping

Source	Output	Type
`.ai/AGENTS.md`	`.opencode/AGENTS.md`	Symlink
`.ai/rules/`	`.opencode/rules/`	Symlink
`.ai/skills/<name>/`	`.opencode/skills/<name>/`	Symlink
`.ai/settings.json`	`opencode.json`	Transformed
`.ai/.opencode/<path>`	`.opencode/<path>`	Symlink

Transformations

Permissions

// Input (Claude format)
{ "permissions": { "allow": ["Bash(git:*)"] } }

// Output (OpenCode format)
{ "permission": { "bash": { "git *": "allow" } } }

MCP Servers

// Input (Claude stdio)
{ "command": "npx", "args": ["-y", "@example/mcp"] }

// Output (OpenCode local)
{ "type": "local", "command": ["npx", "-y", "@example/mcp"] }

Claude	OpenCode
stdio (`command`, `args`)	`type: "local"`, `command: [...]`
`type: "http"` or `"sse"`	`type: "remote"`
`env: {}`	`environment: {}`
`${VAR}`	`{env:VAR}`

Any settings in overrides.opencode are deep-merged into the output.