> For the complete documentation index, see [llms.txt](https://docs.hex-rays.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.hex-rays.com/add-ons/assist/reference/settings-and-commands.md).

# Settings & Commands

This page is a reference for the **Settings** menu in the Assist panel header, the pages inside **Settings → Preferences…**, and the `/` slash commands accepted in the [composer](/add-ons/assist/reference/glossary.md#composer).

## The Settings menu

The panel's **Settings** menu and the **Preferences…** dialog are separate levels — **Sessions**, **Memory**, **Conversation tree…**, **Permissions**, and **Guide…** open their respective windows directly, while **Preferences…** opens the multi-page settings dialog:

```
Settings
├── Sessions
├── Memory
├── Conversation tree…
├── Permissions            (dialog tabs: Assist, MCP clients)
├── Preferences…           (dialog sidebar pages: Agent, Behavior, Appearance,
│                           Sharing & network, Advanced, Developer*, Extensions, Updates)
└── Guide…
```

*\*The **Developer** sidebar page is shown when **Advanced → Developer mode** is **On**.*

| Menu item              | Slash command  | What it opens                                                                                                                                                                                                                                                    |
| ---------------------- | -------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Sessions**           | `/sessions`    | The [saved sessions](/add-ons/assist/reference/glossary.md#saved-session) browser for searching, resuming, exporting, or deleting past conversations and reviewing their recorded write activity (**View write activity**)                                       |
| **Memory**             | `/memory`      | The **Memory** browser for searching, adding, editing, importing/exporting, and reviewing per-binary memories (`NATIVE`, `PENDING`, `INBOUND-PENDING`, and `FOREIGN`; see [Memory](/add-ons/assist/concepts/memory.md))                                          |
| **Conversation tree…** | `/tree`        | The visual conversation tree for inspecting branches, jumping to earlier turns, or forking from a turn *(built-in agent loop only)*                                                                                                                              |
| **Permissions**        | `/permissions` | **Assist permissions** (`Assist` tool-capability tab and `MCP clients` admission tab, where client-admission changes apply immediately; see [Permissions & Privacy](/add-ons/assist/concepts/permissions-and-privacy.md#tool-capability-permissions-assist-tab)) |
| **Preferences…**       | —              | The **Preferences** dialog (individual pages have shortcuts such as `/agent`, `/privacy`, `/sharing`, `/network`, and `/update`; see [Preferences… pages](#preferences-pages) below)                                                                             |
| **Guide…**             | `/guide`       | The interactive in-product Assist guide window (clicking the documentation link inside the guide — or **Help** in the **Preferences** or **Assist permissions** dialogs — opens `docs.hex-rays.com` in your system web browser)                                  |

*(Conversation actions such as `/compact`, `/fulltriage`, and `/export` are run from the composer via* [*slash commands*](#slash-commands) *rather than from the **Settings** menu.)*

## Preferences… pages

Open **Settings → Preferences…** to configure agent backends, panel behavior, appearance, Assist Mesh sharing, advanced thresholds, developer diagnostics, extensions, and update checks. Interactive actions such as account sign-in, adding or removing API keys/endpoints on **Agent**, toggling **Open a network endpoint** on **Sharing & network**, and toggling **Developer mode** on **Advanced** (which reveals the **Developer** sidebar page immediately) act immediately; other settings on these pages take effect when you click **Save**.

### Agent (`/agent`)

Configures the active agent backend, model parameters, per-step CLI timeouts, and standing system prompt instructions (visible controls depend on the selected backend and discovered model capabilities):

| Control                                  | Choices or behavior                                                                                                                                                                                                                                                                  | Notes                                                                                                                                                                                                                                                                                          |
| ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Agent**                                | <p><strong>Assist agent loop:</strong> <code>ChatGPT account</code>, <code>Grok</code>, <code>Claude Code</code>, <code>Antigravity CLI</code>, <code>API key or OpenAI-compatible endpoint</code><br><strong>External agents · separate process:</strong> <code>OpenCode</code></p> | Selects the active agent runtime/backend. See [Getting Started → Choosing an agent](/add-ons/assist/getting-started.md#choose-and-configure-a-model).                                                                                                                                          |
| **Compatible-provider picker**           | Provider/family, **Plan**, **Region** or **Deployment**, **Endpoint**, **Authentication**, named **API key** (**Add key…** / **Save key** / **Remove key**), and **Save endpoint** / **Remove endpoint**                                                                             | Shown when `API key or OpenAI-compatible endpoint` is selected (more than 170 built-in presets across 130+ providers, plus `Custom endpoint...`). If the OS credential store is locked or denied when accessing an API key, Assist reports an error with **Retry** and **Repair permissions**. |
| **Status / sign-in / Account**           | Connection health, sign-in state, and named account selector (**Add account…**, **Remove**)                                                                                                                                                                                          | Shown on account-backed providers (`ChatGPT account`, `Grok`). Consumer account subscriptions and inference API keys are distinct credentials.                                                                                                                                                 |
| **Model**                                | Model list/selection and refresh controls                                                                                                                                                                                                                                            | Populated from the active backend catalogue (or `<base>/models` plus Assist's bundled reference snapshot for OpenAI-compatible endpoints).                                                                                                                                                     |
| **Effort**                               | Model default and supported reasoning levels                                                                                                                                                                                                                                         | Backend- and model-dependent (selected via `-low`/`-high` model variant on Antigravity CLI).                                                                                                                                                                                                   |
| **Speed**                                | Available service-tier choices (e.g., `Standard`, `Fast`)                                                                                                                                                                                                                            | Shown on `ChatGPT account` or when advertised by `OpenCode` over [ACP](/add-ons/assist/reference/glossary.md#acp).                                                                                                                                                                             |
| **Context** *(management)*               | `Automatic` *(working notes, default)* or `Summaries`                                                                                                                                                                                                                                | Shown on `ChatGPT account`.                                                                                                                                                                                                                                                                    |
| **Context** *(token count)*              | Manual token-window capacity spin box (default fallback `131,072`)                                                                                                                                                                                                                   | Shown on `API key or OpenAI-compatible endpoint` when the selected model's context window is not reported by `<base>/models` or the bundled snapshot.                                                                                                                                          |
| **Claude Code → Executable**             | Optional path override                                                                                                                                                                                                                                                               | Shown when `Claude Code` is selected.                                                                                                                                                                                                                                                          |
| **Claude Code → Model step timeout**     | `1`–`60` minutes                                                                                                                                                                                                                                                                     | Timeout for a single model step (not the whole conversation).                                                                                                                                                                                                                                  |
| **Antigravity CLI → Executable**         | Optional `agy` path override                                                                                                                                                                                                                                                         | Shown when `Antigravity CLI` is selected.                                                                                                                                                                                                                                                      |
| **Antigravity CLI → Model step timeout** | `1`–`60` minutes                                                                                                                                                                                                                                                                     | Timeout for a single model step.                                                                                                                                                                                                                                                               |
| **Custom system prompt**                 | Multi-line plain text                                                                                                                                                                                                                                                                | Appends standing instructions to the built-in RE system prompt (updates saved config and reloads the active worker's prompt on **Save**; does not apply to external MCP clients).                                                                                                              |

#### Agent backends compared

| Group                                  | Backend                                   | Sign-in & setup                                         | Accounts / profiles                 | Effort, Speed & Context controls                                          | Conversation rollback, fork, tree, `/fulltriage` & action conversations |
| -------------------------------------- | ----------------------------------------- | ------------------------------------------------------- | ----------------------------------- | ------------------------------------------------------------------------- | ----------------------------------------------------------------------- |
| **Assist agent loop**                  | **ChatGPT account**                       | Browser sign-in                                         | Multiple named accounts             | **Effort**, **Speed**, and **Context** mode                               | ✅                                                                       |
| **Assist agent loop**                  | **Grok**                                  | Browser sign-in                                         | Multiple named accounts             | **Effort** (on supported models); automatic local compaction              | ✅                                                                       |
| **Assist agent loop**                  | **Claude Code**                           | Local `claude` CLI installed and signed in              | Uses active `claude` login          | **Effort** (on supported models); automatic local compaction              | ✅                                                                       |
| **Assist agent loop**                  | **Antigravity CLI**                       | Local `agy` CLI installed and signed in                 | Uses active `agy` login             | Effort selected via model variant; manual `/compact`                      | ✅ *(except auto-compaction)*                                            |
| **Assist agent loop**                  | **API key or OpenAI-compatible endpoint** | 170+ built-in presets or any custom HTTP/HTTPS endpoint | Multiple named endpoints & API keys | **Effort**; configurable **Context** limit when unknown; local compaction | ✅                                                                       |
| **External agents · separate process** | **OpenCode**                              | Local `opencode` CLI installed                          | Configured in OpenCode              | Uses OpenCode-advertised model, effort, and speed options                 | ❌                                                                       |

### Behavior

Controls session restoration, right-click selection actions, proactive suggestion pills, and confidentiality rules:

| Group                              | Control                              | Fresh-profile default / range      | Effect                                                                                                                                                                                                                                                           |
| ---------------------------------- | ------------------------------------ | ---------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Session & navigation**           | **Restore last session on open**     | `On`                               | Reopens the most recent saved conversation for the binary when you open the database.                                                                                                                                                                            |
| **Session & navigation**           | **Navigation**                       | *(Informational)*                  | Describes single-click address navigation and resolvable symbol/address token behavior in the transcript.                                                                                                                                                        |
| **Selection actions**              | **Enable selection actions**         | `On`                               | Populates the **IDA Assist** right-click submenu for selections.                                                                                                                                                                                                 |
| **Selection actions**              | **Auto-run safe read-only actions**  | `Off`                              | Sends qualifying read-only selection actions immediately on click instead of drafting into the composer first.                                                                                                                                                   |
| **Selection actions**              | **Show top action inline**           | `On`                               | Displays the highest-ranked Assist action alongside native IDA popup menu entries.                                                                                                                                                                               |
| **Suggestions**                    | **Show suggestions**                 | `On`                               | Displays the context-sensitive [suggestion pill](/add-ons/assist/reference/glossary.md#suggestion-pill) strip above the composer.                                                                                                                                |
| **Suggestions**                    | **Let the model suggest follow-ups** | `On`                               | Allows model-proposed follow-up suggestions.                                                                                                                                                                                                                     |
| **Suggestions**                    | **Maximum shown**                    | `4` *(range `0`–`12`)*             | Maximum suggestion pills shown at once (`0` hides displayed pills while keeping generation active).                                                                                                                                                              |
| **Privacy & context** (`/privacy`) | **Confidentiality rules / Rule set** | No active rules in a fresh profile | Configures outbound/inbound text redaction rules (see [Confidentiality rule editor](#confidentiality-rule-editor-privacy) below and [Permissions & Privacy → Confidentiality rules](/add-ons/assist/concepts/permissions-and-privacy.md#confidentiality-rules)). |

#### Confidentiality rule editor (`/privacy`)

The **Confidentiality rules** editor under **Behavior → Privacy & context** (`/privacy`) manages global and per-binary redaction rules:

| Control                                | Choices or behavior                                                                                                              |
| -------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| **Rule set**                           | `User defaults` or `This binary` (available when the open database has a binary identity)                                        |
| **Rule table columns**                 | `On`, `Apply to`, `Match type`, `Original`, `Replacement`, `Ignore case`, `Replace`                                              |
| **Apply to**                           | `All text`, `Function names`, `Binary name`, `Variable names`, `String literals`                                                 |
| **Match type**                         | `Literal` or `Regex` (RE2 syntax; regex replacements support `$1`–`$9` and `$$`, while literal replacements are copied verbatim) |
| **Replace**                            | `Matched text` or `Whole value` (`Whole value` with `All text` can refuse when a match lies outside a known containing value)    |
| **Add**                                | Appends a disabled row until a valid rule is configured                                                                          |
| **Remove / disable**                   | Removes a local row or disables an inherited row according to scope                                                              |
| **Move up / Move down**                | Reorders rows for tie-breaking                                                                                                   |
| **Inherit selected**                   | Removes a selected binary override in favor of `User defaults` inheritance                                                       |
| **Local preview text**                 | Local test input showing the projected result and any shadowing or validation errors                                             |
| **Reset invalid replacement settings** | Recovery button shown if stored rule configuration is invalid                                                                    |

Rules are evaluated in deterministic precedence order: **binary overrides first**, then **category specificity**, then **longest match**, and finally **table row order** for ties.

### Appearance

Adjusts color contrast and transcript typography (Assist follows IDA's active theme):

| Control               | Choices / range                                                                | Notes                                                                                 |
| --------------------- | ------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------- |
| **Color readability** | `Auto-adjust colors for readability` *(default)* or `Use theme colors exactly` | Controls whether Assist adjusts palette contrast for readability against IDA's theme. |
| **Font family**       | `Default` or any installed font family                                         | Applies to prose in the panel; code blocks remain monospace.                          |
| **Font size**         | `Match IDA` or `8`–`18 pt`                                                     | Point size for new messages (reopen the panel to restyle the entire open transcript). |
| **Preview**           | Pseudocode sample                                                              | Informational preview showing how current font and color-readability choices render.  |

### Sharing & network (`/sharing`, `/network`)

Configures local Assist Mesh linking, cross-database memory sharing, external MCP client registration snippets, and the temporary TCP network endpoint (see [Assist Mesh → Mesh settings](/add-ons/assist/concepts/mesh.md#mesh-settings)):

| Group                             | Control                                                              | Default / behavior                                                                                                                                                                                                                          |
| --------------------------------- | -------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Mesh**                          | **Keep linked when chat is closed**                                  | `On` — keeps the local linked service active for peer IDA instances and external MCP clients when the Assist panel is closed.                                                                                                               |
| **Mesh**                          | **Send memories**                                                    | `Confirm each send, and let the model suggest them` *(default)*, `Confirm each send`, or `Send automatically for this binary` (controls outbound `propagate_memory` calls; inbound peer memories are always held in quarantine for review). |
| **Connected nodes**               | **Manage linked databases**                                          | Displays currently linked peer databases and opens linked-database management.                                                                                                                                                              |
| **Mesh MCP**                      | **Claude Code**, **Codex**, **Grok**, **Antigravity CLI** (**Copy**) | Four generated CLI registration commands pointing to the bundled `assist --mcp` bridge.                                                                                                                                                     |
| **Mesh MCP**                      | **Other agents — `mcp.json` or `config.json`** (**Copy**)            | Generic `"ida"` MCP server entry block for other MCP clients.                                                                                                                                                                               |
| **Network endpoint** (`/network`) | **Open a network endpoint**                                          | `Off` — live, temporary toggle (takes effect immediately; **Cancel** does not revert it, and it is never persisted across restarts).                                                                                                        |
| **Network endpoint**              | **Listen on**                                                        | Select from available local interfaces, `Every address`, loopback (`127.0.0.1`), or a typed address.                                                                                                                                        |
| **Network endpoint**              | **Address in command**                                               | Shown conditionally when the client command needs a separate dial address (IPv6 endpoints format as `[address]:port`).                                                                                                                      |
| **Network endpoint**              | **Command / JSON `Copy`**                                            | Copies the full client configuration including `ASSIST_MCP_CONNECTION_CODE` (masked in the visible preview; valid for that active endpoint generation).                                                                                     |

### Advanced

Controls multi-select function batch behavior, Imports/Exports AI menu items, Developer mode visibility, the resolved data directory, and persistent-memory master switches:

| Group                         | Control                                    | Default                  | Range / behavior                                                                                                                                                                                |
| ----------------------------- | ------------------------------------------ | ------------------------ | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Group actions**             | **Confirm above**                          | `150` functions          | `1`–`100,000` (actions that modify the database also prompt for confirmation).                                                                                                                  |
| **Group actions**             | **Direct right-click actions**             | `7`                      | `0`–`64`; additional actions overflow under **More Assist actions**.                                                                                                                            |
| **Group actions**             | **Timeout**                                | `30,000 ms`              | `1,000`–`600,000 ms`; checked between per-function items rather than preempting an in-flight native IDA call.                                                                                   |
| **Group actions**             | **Page size**                              | `50` rows                | `1`–`250` rows per result page.                                                                                                                                                                 |
| **Group actions**             | **Group-action toggles (10)**              | All `On`                 | Enables or disables individual multi-select function actions in the Functions window.                                                                                                           |
| **Imports & Exports windows** | **Allow AI actions on selected imports**   | `On`                     | Shows model-assisted actions on selected imports (local xref lookups remain available when `Off`).                                                                                              |
| **Imports & Exports windows** | **Allow AI actions on selected exports**   | `On`                     | Shows model-assisted actions on selected exports (local caller/reachability lookups remain available when `Off`).                                                                               |
| **Developer**                 | **Developer mode**                         | `Off`                    | Reveals the **Developer** page in the **Preferences** sidebar immediately when toggled, and enables internal result/detail affordances on **Save**.                                             |
| **Data directory**            | **Path** / **Copy path** / **Open folder** | `<user-ida-dir>/assist/` | Read-only display of the active writable data root (resolved from `ASSIST_HOME` or `IDAUSR`; see [Diagnostics & Data Storage](/add-ons/assist/reference/diagnostics.md#environment-variables)). |
| **Persistent memory**         | **Allow memory recall**                    | `On`                     | Master switch for read-only memory retrieval (`recall`).                                                                                                                                        |
| **Persistent memory**         | **Allow memory writes**                    | `On`                     | Master switch for new or revised memories (external proposals still require review).                                                                                                            |
| **Persistent memory**         | **Include memory in prompts**              | `On`                     | Automatically injects up to 6 local and 6 global accepted memories into each turn's prompt context.                                                                                             |
| **Persistent memory**         | **Harvest memory during compaction**       | `On`                     | Allows `/compact` to stage durable memory proposals for commit through the memory write gate.                                                                                                   |

### Developer

Visible in the **Preferences** sidebar only when **Advanced → Developer mode** is enabled. Provides agent-loop compaction and execution budgets, opt-in per-session diagnostic logging channels, and read-only token/field usage reports (see [Diagnostics & Data Storage → Diagnostic logs and Developer mode](/add-ons/assist/reference/diagnostics.md#diagnostic-logs-and-developer-mode)):

| Group                     | Control                               | Default / range                                                     | Meaning                                                                                                                                                                                           |
| ------------------------- | ------------------------------------- | ------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Agent loop**            | **Compact tool results for the AI**   | `On`                                                                | Compacts model-facing tool outputs while preserving full analyst rendering in the transcript.                                                                                                     |
| **Agent loop**            | **AI copy profile**                   | `Balanced` (`Aggressive`, `Balanced`, `Verbatim`)                   | Controls the detail density of model-facing tool results.                                                                                                                                         |
| **Agent loop**            | **Step limit**                        | `0` / `Unlimited` (`0`–`500`)                                       | Maximum model steps per user message (`0` = unlimited).                                                                                                                                           |
| **Agent loop**            | **Script limit**                      | `5 s` (`1`–`600 s`)                                                 | Advisory timeout budget for `idascript` (cannot preempt native decompilation or auto-analysis inside IDA).                                                                                        |
| **Agent loop**            | **Learn locally from action choices** | `On`                                                                | Writes diagnostic counters to separate local telemetry storage and excludes them from memory recall (never uploaded).                                                                             |
| **Debug logging**         | **External agent traffic**            | `Off`                                                               | Logs raw wire messages exchanged with Claude Code, OpenCode, and other external agents (`logs/sessions/<session>.jsonl`).                                                                         |
| **Debug logging**         | **Built-in agent requests**           | `Off`                                                               | Logs full model API requests, returned items, token usage, and errors (`logs/sessions/<session>.jsonl`).                                                                                          |
| **Debug logging**         | **Agent turns and tools (Perfetto)**  | `Off`                                                               | Writes Perfetto timing and queue profiles (`profiles/native/*.json`; payloads omitted).                                                                                                           |
| **Debug logging**         | **Executed IDA scripts**              | `Off`                                                               | Logs every `idascript` Assist runs along with its outcome.                                                                                                                                        |
| **Debug logging**         | **Suggestion-engine diagnostics**     | `Off`                                                               | Logs why suggestion pills were refreshed, ranked, shown, or skipped.                                                                                                                              |
| **Debug logging**         | **Tool calls and results only**       | `Off`                                                               | Narrows enabled agent session logs to tool invocations and outputs, omitting full prompt/request payloads.                                                                                        |
| **Debug logging**         | **Logs / Profiles / Open folder**     | `<assist-home>/logs/sessions/` and `<assist-home>/profiles/native/` | Displays computed session-log and Perfetto-profile paths and opens the folder in your system file manager *(note: the always-on `logs/assist_debug.log` is separate from these opt-in channels)*. |
| **Read-only diagnostics** | **AI copy token report**              | *(Session table)*                                                   | Displays per-session token measurements with **Show raw report** / **Hide raw report**.                                                                                                           |
| **Read-only diagnostics** | **AI copy field use**                 | *(Advisory report)*                                                 | Displays field-level usage in model-facing results (does not change result rendering).                                                                                                            |

### Extensions

Displays installed local extension cards from `<user-ida-dir>/assist/extensions/` along with state-dependent management controls:

* **Extension cards & state controls** — depending on each extension's state and capabilities, cards expose **Trust…**, **Enable**, **Disable**, **Reload**, **Re-sync now**, **View log**, **Remove**, **Copy error**, and (on token-bearing cards) **Copy** / **Rotate**, plus filter-chain ordering when multiple filter extensions are present.
* **Trust & permission model** — extensions run under the intersection of their granted extension trust scopes and Assist's host capability permissions (not an unconditional grant).

### Updates (`/update`)

Configures release update checks against `hexrays.plugins.hex-rays.com`:

| Control                             | Behavior                                                                                                                                             |
| ----------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Check for updates automatically** | `On` by default; checks anonymously shortly after IDA startup and once an hour while IDA is open (nothing is downloaded or installed automatically). |
| **Check action**                    | Runs an immediate manual check against the plugin portal (label reflects current check state).                                                       |
| **Stop**                            | Cancels an in-flight manual update check.                                                                                                            |
| **Copy upgrade command**            | Shown when a newer release is available; copies `hcli plugin upgrade hexrays/assist`.                                                                |
| **Release / status view**           | Displays the installed version and latest available release status.                                                                                  |

## Slash commands

Type `/` at the start of the composer to run a panel command. Unrecognized `/...` strings are sent to the active agent as ordinary message text:

| Command        | Action / availability                                                                                                                                     |
| -------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `/agent`       | Open **Settings → Preferences… → Agent**                                                                                                                  |
| `/compact`     | Compress active built-in conversation history into a local summary checkpoint *(built-in agent loop; requires conversation messages and no running turn)* |
| `/export`      | Save the visible conversation as an HTML transcript *(requires conversation messages; does not require a connected model)*                                |
| `/fulltriage`  | Launch full binary triage and open a **Triage** conversation *(built-in agent loop on a database with an input file digest)*                              |
| `/guide`       | Open the interactive Assist guide window (**Settings → Guide…**)                                                                                          |
| `/memory`      | Open the **Memory** browser (**Settings → Memory**)                                                                                                       |
| `/network`     | Open **Settings → Preferences… → Sharing & network**                                                                                                      |
| `/permissions` | Open **Settings → Permissions** (`Assist` and `MCP clients` tabs)                                                                                         |
| `/privacy`     | Open **Settings → Preferences… → Behavior → Privacy & context** (**Confidentiality rules**)                                                               |
| `/sessions`    | Open the [saved sessions](/add-ons/assist/reference/glossary.md#saved-session) browser (**Settings → Sessions**)                                          |
| `/sharing`     | Open **Settings → Preferences… → Sharing & network**                                                                                                      |
| `/tree`        | Open the visual conversation tree (**Settings → Conversation tree…**; built-in agent loop only)                                                           |
| `/update`      | Open **Settings → Preferences… → Updates**                                                                                                                |


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.hex-rays.com/add-ons/assist/reference/settings-and-commands.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.