# Augment

> Import Auggie CLI, Augment IDE, Cosmos sessions, instructions, and local memories.

Hansard imports Augment's local histories and explicitly requested Cosmos
cloud sessions. Recorded models retain their own maker in Stats.

## Source IDs

| Source          | Use it for                                                                  |
| --------------- | --------------------------------------------------------------------------- |
| `augment`       | Auggie JSON sessions, Augment IDE stores, and existing Cosmos local caches. |
| `augment-cloud` | Read-only Cosmos API history, including child sessions.                     |

## How the import works

Auggie stores completed and in-progress exchanges under `~/.augment/sessions`.
Hansard preserves the original files, maps tool calls and results, and uses
recorded file contents for numbered edit diffs. Pending queue entries do not
count as submitted messages. Resuming a session updates its existing archive.
Sync conflict backups stay with the session's raw files; their copied history
does not add duplicate messages or usage.

The IDE stores exchanges and tool states in LevelDB, with an ordered history
index and optional conversation metadata in saved webview state. Hansard takes
a stable physical snapshot before opening LevelDB, leaving the original store
and its lock with the provider. Missing pointers and conflicting copies are
reported as import errors.

Cosmos imports follow listing pages, fetch child sessions, and compare each
download against the server's message count before accepting it. An existing
complete cloud archive takes precedence over the CLI's shorter local sync.
The importer does not start agents, publish session links, or invoke sync.

## Import and keep current

**Conversation history**

```sh
hansard import --source augment --since all
hansard config sources enable augment
hansard import augment-cloud
```

Cloud import uses the existing `auggie login` session or the documented
`AUGMENT_SESSION_AUTH` environment variable. Credential files are not archived.
Only HTTPS Augment tenant URLs are accepted. `--session <agent-id>` selects one
cloud session and its children. Turn on **Augment cloud agents** in **Settings →
Imports → Cloud agents** to have the watcher refresh them every six hours.

Set `HANSARD_AUGMENT_ROOTS` to custom Auggie cache roots, including a directory
used with `--augment-cache-dir`. `HANSARD_AUGMENT_IDE_ROOTS` selects Code-family
`User` directories. Both accept path-separated lists. Default IDE discovery
covers supported Code-family application and remote-server storage locations.

**Memories and instructions**

```sh
hansard memory backup
```

Backups include global and project guidelines, rules, commands, agent
Markdown, and migrated context. Recognizable memory directories and memory,
knowledge, and learnings Markdown files in local Cosmos VFS mirrors are also
captured. `AUGMENT_VFS_DIR` selects a custom mirror. Skill folders, unrelated
VFS documents, credentials, connector configuration, and remote-only memories
are excluded.

## Review the import in the app

1. When the import finishes, open **Conversations** and select **Filter & sort**.
2. Set **Source** to **Augment** and set **Time** to the window you imported. Use **Folder** to narrow the list when you use this harness in several projects.
3. Select a conversation to check its messages, tool calls, files, commits, model, and token usage.

To keep archiving new sessions from a local harness, open **Settings → Imports**. **Automatic** follows every supported harness installed on this computer. **Custom** lets you turn individual harnesses on or off. Cloud sources and account exports never run from here; each needs its own import command. If a conversation you expect is missing after you reset the filters, run the import command on this page again with `--explain-skips` where the source supports it.

Check tool arguments, results, and numbered edits. In Stats, compare recorded
tokens and models; credits are separate metadata. Open Memories to inspect
instruction snapshots.

## What Hansard preserves

- Prompts, replies, readable reasoning, and embedded image or file attachments.
- Tool inputs and results, MCP attribution, and available historical edits.
- Embedded subagent runs and recorded parent relationships.
- Recorded token counters, billing metadata, and model identities.
- Native CLI backups, coherent IDE database files, and cloud API records.

## Known limitations

- Native verification covers CLI listing of a synthetic session and isolated
  imports. Authenticated model calls, native delegation, and live IDE edits
  remain pending verification.
- CLI message clocks often mark exchange completion. Unknown record types are
  counted in metadata and retained in raw files.
- The native Cosmos sync can retain only 200 messages. Use `augment-cloud` for
  a count-verified download; truncated or changing responses require a retry.
- Cloud transcript records may omit models and token usage. The audited cost
  field is an estimate. Child-inclusive cost totals are retained without adding
  them to the parent's accounting.
- IDE state may have lost older message bodies. Missing indexed exchanges
  remain an error rather than an inferred transcript.
- Conflict backups are raw-only. An orphan backup requires restoration through
  Augment before import. Lossy migrated thread summaries do not replace full
  conversation histories.
- Expert memories with custom filenames outside recognizable local memory
  paths, and cloud memories without a local mirror, are not discovered.

## Refresh an existing archive

```sh
hansard import --source augment --since all
hansard import augment-cloud
hansard memory backup
```

Keep original stores for parser refreshes. Local CLI conversations offer a
resume command using their recorded session ID and cache directory.

Official references: [CLI reference](https://docs.augmentcode.com/cli/reference),
[cloud commands](https://docs.augmentcode.com/cli/cloud), and
[Expert memory](https://docs.augmentcode.com/cosmos/experts-memory).