opencode-transcript

Create clean Markdown transcripts from OpenCode sessions.

Each transcript contains only genuine user messages, user attachments, and final assistant responses. Tool calls and results, reasoning, progress text, compaction messages, synthetic prompts, and child agent sessions are excluded. Switching between plan and build agents does not split or label the transcript.

Output

.opencode/transcripts/<session-title>-<created>/
|-- transcript.md
`-- references/

The plugin waits until the session has a title before creating its transcript.

Pasted images are copied into references/ and embedded in the transcript. Other pasted files are copied and linked. Reverted messages and their unused references disappear from the generated transcript; unreverting restores them while OpenCode still retains the session data.

To enable transcript creation, add .opencode/transcripts/ to the worktree's root .gitignore. The plugin does not modify .gitignore and remains inactive until that entry exists.

Install From GitHub

Clone and build the plugin:

git clone [email protected]:alizeait/opencode-transcript.git
cd opencode-transcript
pnpm install
pnpm build

Add the absolute repository directory to ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["file:///absolute/path/to/opencode-transcript"]
}

On Windows, use a file URL such as file:///C:/Users/you/src/opencode-transcript.

Quit and restart OpenCode after installing or rebuilding the plugin. OpenCode loads plugin code only at startup.

Install From npm

pnpm add opencode-transcript

Then add the package to ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-transcript"]
}

Development

pnpm install
pnpm test
pnpm check-types
pnpm lint
pnpm format:check
pnpm build