autter attributes a line to an AI agent when the agent tells autter, right before and right after it edits a file. Each agent below does that in one of three ways:
- CLI hooks:
autter install-hooksadds hook commands to the agent's own config file. They runautter checkpoint <preset> --hook-input stdin. - Plugin:
autter install-hookswrites a plugin file the agent loads, and the plugin runsautter checkpoint. - IDE extension: an editor extension or IDE plugin watches edits and sends the checkpoints itself.
autter install-hooks (or autter onboard) sets up every agent it finds on the
machine. To check one agent, run its autter doctor --agent command. It reports
whether the agent is detected, which files autter manages for it, whether each
hook command runs, and then sends the agent's own pre- and post-edit hook calls
through autter in a throwaway repository. It passes when the background service
records an AI checkpoint for the edited file, so no real agent session is needed.
Every failed step prints a fix: line.
| Agent | Integration | What is captured | What is not | Check it |
|---|---|---|---|---|
| Claude Code | CLI hooks in ~/.claude/settings.json (and Claude Agent SDK harness settings) |
Every tool edit (Write, Edit, MultiEdit, any tool that names a file) and Bash commands; model and prompts from the transcript | Claude Code older than 2.0 | autter doctor --agent claude-code |
| Codex | CLI hooks in ~/.codex/config.toml (or ~/.codex/hooks.json) |
apply_patch edits and shell commands; prompts from the rollout file |
The model when Codex's hook does not send it (recorded as unknown) | autter doctor --agent codex |
| Cursor | CLI hooks in ~/.cursor/hooks.json, plus the autter extension |
Write, StrReplace and Delete edits and Shell commands from the agent; model from the hook or transcript | Tab completions (only with the extension's experimental AI-tab tracking); Cursor older than 1.7 | autter doctor --agent cursor |
| GitHub Copilot (VS Code agent mode and Copilot CLI) | CLI hooks in ~/.copilot/hooks/autter.json |
File edits and terminal commands in agent mode and the Copilot CLI; model from the chat session | Inline completions and inline chat; VS Code older than 1.99 | autter doctor --agent github-copilot |
| Amp | Plugin ~/.config/amp/plugins/autter.ts |
Edit and Write tool calls and Bash commands that finish; model from the thread | Failed or cancelled tool calls | autter doctor --agent amp |
| OpenCode | Plugin ~/.config/opencode/plugins/autter.ts |
edit, write, patch, multiedit and apply_patch, plus bash; model from OpenCode's store | Other tools | autter doctor --agent opencode |
| Pi | Plugin ~/.pi/agent/extensions/autter.ts |
edit and write tool calls and bash (more tools through ~/.pi/agent/autter.override.json); model from the session file |
Tools not in the policy | autter doctor --agent pi |
| Gemini CLI | CLI hooks in ~/.gemini/settings.json |
write_file and replace edits and shell commands; model from the transcript | autter doctor --agent gemini |
|
| Droid (Factory) | CLI hooks in ~/.factory/settings.json |
ApplyPatch, Edit, Write and Create edits and Bash commands | autter doctor --agent droid |
|
| Firebender | CLI hooks in ~/.firebender/hooks.json |
Write, Edit, Delete and symbol edits and Bash commands; model from the hook | autter doctor --agent firebender |
|
| Google Antigravity | CLI hooks in ~/.gemini/config/hooks.json |
File writes and run_command in the Antigravity CLI and 2.0 app |
The Antigravity IDE has been reported not to run hooks (details) | autter doctor --agent antigravity |
| Kiro | CLI hooks in ~/.kiro/hooks/autter.json (Kiro CLI v3) |
File writes and shell commands from the Kiro CLI agent | Kiro CLI 2.x; the Kiro IDE unless it loads ~/.kiro/hooks/; the model and prompts. Payload fields not yet confirmed (details) |
autter doctor --agent kiro |
| Windsurf | CLI hooks in ~/.codeium/windsurf/hooks.json (and ~/.codeium/hooks.json), plus the autter extension |
Cascade file writes and commands | Tab completions | autter doctor --agent windsurf |
| VS Code | IDE extension autter.autter-vscode |
Your own typing (known human); Copilot edits on VS Code older than 1.109.3, where native hooks are missing | Inline chat; AI tab completions unless the experimental setting is on | autter doctor --agent vscode |
| JetBrains IDEs | IDE plugin (build 252 or newer) | Edits by GitHub Copilot for JetBrains and Junie | The model and prompts; other JetBrains AI plugins | autter doctor --agent jetbrains |
| Visual Studio (Windows) | IDE extension (VSIX, Visual Studio 2022) | Copilot edits | The model and prompts | autter doctor --agent visual-studio |
Token usage, including prompt-cache reads and writes, is described per agent in agent-usage.md.
Shell commands an agent runs (Bash, run_in_terminal, ...) are captured by
comparing the files before and after the command. Gitignored files and deletions
are not attributed; see shell-edit-capture.md.
For an editor extension there is no hook call to simulate:
autter doctor --agent checks that the extension is installed, and the simulated
edit is skipped. To check one, edit a file in the editor, then run autter status.
Agents not in this table have no autter integration yet. Their edits show as
untracked in autter stats and as yours in autter blame;
unsupported-agents.md lists the ones autter doctor
recognises and what is missing. To add an agent, see the
agent integration guide.