> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rippletide.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Privacy & capture

> Choose what runtime content to export and distinguish traces, inventory and decision audit data.

Trace capture, inventory declaration and policy evaluation are separate data paths. Reducing trace capture does not prevent a policy check from receiving the values it must evaluate.

## Defaults by integration

| Integration                                      | Default trace behavior                                                | Content controls                                                                  |
| ------------------------------------------------ | --------------------------------------------------------------------- | --------------------------------------------------------------------------------- |
| TypeScript SDK, including instrumented JS/TS MCP | `full`: bounded payload values with credential-shaped values redacted | `captureMode` client option or `RIPPLETIDE_CAPTURE_MODE`                          |
| Python SDK                                       | `metadata`                                                            | `RIPPLETIDE_CAPTURE_MODE`; prompt text needs separate `allow_content_export=True` |
| Codex live hooks                                 | Activity storage off until selected; `redacted` when enabled          | Setup `--traces`/`--no-traces`, then capture setting in `~/.rippletide/codex.env` |
| Codex history import                             | Explicit import, `redacted` by default                                | `backfill --capture metadata`, `redacted`, or `full`                              |

Do not assume a metadata-only default across all SDKs. Set the capture mode deliberately before running an integration on customer content.

## Capture modes

| Mode       | Payloads                                                                                                   |
| ---------- | ---------------------------------------------------------------------------------------------------------- |
| `metadata` | Timing, kinds, correlation identifiers, model/token and decision metadata; input/output payloads withheld  |
| `redacted` | Payload structure, with scalar values replaced by redaction markers                                        |
| `full`     | Bounded input/output values; credential-shaped keys and authorization values receive best-effort redaction |

For an SDK agent, place a setting in its environment:

```bash theme={null}
RIPPLETIDE_CAPTURE_MODE=metadata
```

For Codex, put it in `~/.rippletide/codex.env`; repository environment variables do not configure that adapter. Use supported mode names exactly.

Metadata can still be sensitive. Identifiers, tool names, model names and attributes may identify a person or system. Codex metadata can include an absolute working directory containing a username. Redaction is not a guarantee that all personal or business-sensitive data has been removed.

## Inventory and prompts

Third-party connection declarations send credential names and status, not secret values. Prompt/instruction inventory is distinct from trace payload capture: a TypeScript prompt declaration can send its fallback text even when runtime capture is metadata. Review what you register.

Python withholds prompt text by default. Enable `allow_content_export=True` only when the prompt owner has approved storing it. Do not put secrets in prompt text or schema descriptions.

## Policy decision data

Before-action parameters and trusted context are sent to Rippletide and stored in the decision audit trail. Before-response checks evaluate the candidate response and request. These policy inputs do not inherit the trace redaction setting. Keep their scope limited to facts needed by the policy.

Result filtering sends the full declared result for transient validation and evaluation. Raw collection/item bodies are not persisted in the result decision; item IDs, fingerprints, decisions, reasons and plan metadata are retained. Python requires explicit `allow_result_export=True`. Runtime traces can separately capture result content according to their capture mode.

Rules using a model predicate can send the selected field and instruction to the configured model evaluator. Trace discovery analyses captured values, and document discovery/labeling analyses imported text. Only select sources and fields approved for that processing. A metadata-only trace cannot later be replayed as if its raw arguments had been captured.

## Codex history and local state

Backfill reads local Codex rollouts only when explicitly requested, or when you opt into history import during setup. Use `--dry-run --stats` first. `full` can export prompts, patches, command lines and command output; the history does not receive retroactive policy decisions.

Removing local instrumentation or hooks does not erase previously stored server evidence. Keep Connection keys and local credential files private; use [API keys](/docs/api-keys) for rotation and revocation.