agent-approvals-security diff

agent-approvals-security.md +330 −0 added

Details

1# Agent approvals & security

3Codex helps protect your code and data and reduces the risk of misuse.

5This page covers how to operate Codex safely, including sandboxing, approvals,

6 and network access. If you are looking for Codex Security, the product for

7 scanning connected GitHub repositories, see [Codex Security](https://developers.openai.com/codex/security).

9By default, the agent runs with network access turned off. Locally, Codex uses an OS-enforced sandbox that limits what it can touch (typically to the current workspace), plus an approval policy that controls when it must stop and ask you before acting.

11For a high-level explanation of how sandboxing works across the Codex app, IDE

12extension, and CLI, see [sandboxing](https://developers.openai.com/codex/concepts/sandboxing).

13For a broader enterprise security overview, see the [Codex security white paper](https://trust.openai.com/?itemUid=382f924d-54f3-43a8-a9df-c39e6c959958&source=click).

15## Sandbox and approvals

17Codex security controls come from two layers that work together:

19- **Sandbox mode**: What Codex can do technically (for example, where it can write and whether it can reach the network) when it executes model-generated commands.

20- **Approval policy**: When Codex must ask you before it executes an action (for example, leaving the sandbox, using the network, or running commands outside a trusted set).

22Codex uses different sandbox modes depending on where you run it:

24- **Codex cloud**: Runs in isolated OpenAI-managed containers, preventing access to your host system or unrelated data. Uses a two-phase runtime model: setup runs before the agent phase and can access the network to install specified dependencies, then the agent phase runs offline by default unless you enable internet access for that environment. Secrets configured for cloud environments are available only during setup and are removed before the agent phase starts.

25- **Codex CLI / IDE extension**: OS-level mechanisms enforce sandbox policies. Defaults include no network access and write permissions limited to the active workspace. You can configure the sandbox, approval policy, and network settings based on your risk tolerance.

27In the `Auto` preset (for example, `--full-auto`), Codex can read files, make edits, and run commands in the working directory automatically.

29Codex asks for approval to edit files outside the workspace or to run commands that require network access. If you want to chat or plan without making changes, switch to `read-only` mode with the `/permissions` command.

31Codex can also elicit approval for app (connector) tool calls that advertise side effects, even when the action isn't a shell command or file change. Destructive app/MCP tool calls always require approval when the tool advertises a destructive annotation, even if it also advertises other hints (for example, read-only hints).

33## Network access [Elevated Risk](https://help.openai.com/articles/20001061)

35For Codex cloud, see [agent internet access](https://developers.openai.com/codex/cloud/internet-access) to enable full internet access or a domain allow list.

37For the Codex app, CLI, or IDE Extension, the default `workspace-write` sandbox mode keeps network access turned off unless you enable it in your configuration:

39```toml

40[sandbox_workspace_write]

41network_access = true

42```

44You can also control the [web search tool](https://platform.openai.com/docs/guides/tools-web-search) without granting full network access to spawned commands. Codex defaults to using a web search cache to access results. The cache is an OpenAI-maintained index of web results, so cached mode returns pre-indexed results instead of fetching live pages. This reduces exposure to prompt injection from arbitrary live content, but you should still treat web results as untrusted. If you are using `--yolo` or another [full access sandbox setting](#common-sandbox-and-approval-combinations), web search defaults to live results. Use `--search` or set `web_search = "live"` to allow live browsing, or set it to `"disabled"` to turn the tool off:

46```toml

47web_search = "cached" # default

48# web_search = "disabled"

49# web_search = "live" # same as --search

50```

52Use caution when enabling network access or web search in Codex. Prompt injection can cause the agent to fetch and follow untrusted instructions.

54## Defaults and recommendations

56- On launch, Codex detects whether the folder is version-controlled and recommends:

57 - Version-controlled folders: `Auto` (workspace write + on-request approvals)

58 - Non-version-controlled folders: `read-only`

59- Depending on your setup, Codex may also start in `read-only` until you explicitly trust the working directory (for example, via an onboarding prompt or `/permissions`).

60- The workspace includes the current directory and temporary directories like `/tmp`. Use the `/status` command to see which directories are in the workspace.

61- To accept the defaults, run `codex`.

62- You can set these explicitly:

63 - `codex --sandbox workspace-write --ask-for-approval on-request`

64 - `codex --sandbox read-only --ask-for-approval on-request`

66### Protected paths in writable roots

68In the default `workspace-write` sandbox policy, writable roots still include protected paths:

70- `<writable_root>/.git` is protected as read-only whether it appears as a directory or file.

71- If `<writable_root>/.git` is a pointer file (`gitdir: ...`), the resolved Git directory path is also protected as read-only.

72- `<writable_root>/.agents` is protected as read-only when it exists as a directory.

73- `<writable_root>/.codex` is protected as read-only when it exists as a directory.

74- Protection is recursive, so everything under those paths is read-only.

76### Deny reads with filesystem profiles

78Named permission profiles can also deny reads for exact paths or glob patterns.

79This is useful when a workspace should stay writable but specific sensitive

80files, such as local environment files, must stay unreadable:

82```toml

83default_permissions = "workspace"

85[permissions.workspace.filesystem]

86":project_roots" = { "." = "write", "**/*.env" = "none" }

87glob_scan_max_depth = 3

88```

90Use `"none"` for paths or globs that Codex shouldn't read. The sandbox policy

91evaluates globs for local macOS and Linux command execution. On platforms that

92pre-expand glob matches before the sandbox starts, set `glob_scan_max_depth` for

93unbounded `**` patterns, or list explicit depths such as `*.env`, `*/*.env`, and

94`*/*/*.env`.

96### Run without approval prompts

98You can disable approval prompts with `--ask-for-approval never` or `-a never` (shorthand).

100This option works with all `--sandbox` modes, so you still control Codex's level of autonomy. Codex makes a best effort within the constraints you set.

101

102If you need Codex to read files, make edits, and run commands with network access without approval prompts, use `--sandbox danger-full-access` (or the `--dangerously-bypass-approvals-and-sandbox` flag). Use caution before doing so.

103

104For a middle ground, `approval_policy = { granular = { ... } }` lets you keep specific approval prompt categories interactive while automatically rejecting others. The granular policy covers sandbox approvals, execpolicy-rule prompts, MCP prompts, `request_permissions` prompts, and skill-script approvals.

105

106Set `approvals_reviewer = "guardian_subagent"` to route eligible approval reviews through the Guardian reviewer subagent instead of prompting the user directly. Admin requirements can constrain this with `allowed_approvals_reviewers`.

107

108### Common sandbox and approval combinations

109

110| Intent | Flags | Effect |

111| ----------------------------------------------------------------- | ------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------ |

112| Auto (preset) | *no flags needed* or `--full-auto` | Codex can read files, make edits, and run commands in the workspace. Codex requires approval to edit outside the workspace or to access network. |

113| Safe read-only browsing | `--sandbox read-only --ask-for-approval on-request` | Codex can read files and answer questions. Codex requires approval to make edits, run commands, or access network. |

114| Read-only non-interactive (CI) | `--sandbox read-only --ask-for-approval never` | Codex can only read files; never asks for approval. |

115| Automatically edit but ask for approval to run untrusted commands | `--sandbox workspace-write --ask-for-approval untrusted` | Codex can read and edit files but asks for approval before running untrusted commands. |

116| Dangerous full access | `--dangerously-bypass-approvals-and-sandbox` (alias: `--yolo`) | [Elevated Risk](https://help.openai.com/articles/20001061) No sandbox; no approvals *(not recommended)* |

117

118`--full-auto` is a convenience alias for `--sandbox workspace-write --ask-for-approval on-request`.

119

120With `--ask-for-approval untrusted`, Codex runs only known-safe read operations automatically. Commands that can mutate state or trigger external execution paths (for example, destructive Git operations or Git output/config-override flags) require approval.

121

122#### Configuration in `config.toml`

123

124For the broader configuration workflow, see [Config basics](https://developers.openai.com/codex/config-basic), [Advanced Config](https://developers.openai.com/codex/config-advanced#approval-policies-and-sandbox-modes), and the [Configuration Reference](https://developers.openai.com/codex/config-reference).

125

126```toml

127# Always ask for approval mode

128approval_policy = "untrusted"

129sandbox_mode = "read-only"

130allow_login_shell = false # optional hardening: disallow login shells for shell-based tools

131

132# Optional: Allow network in workspace-write mode

133[sandbox_workspace_write]

134network_access = true

135

136# Optional: granular approval policy

137# approval_policy = { granular = {

138# sandbox_approval = true,

139# rules = true,

140# mcp_elicitations = true,

141# request_permissions = false,

142# skill_approval = false

143# } }

144```

145

146You can also save presets as profiles, then select them with `codex --profile <name>`:

147

148```toml

149[profiles.full_auto]

150approval_policy = "on-request"

151sandbox_mode = "workspace-write"

152

153[profiles.readonly_quiet]

154approval_policy = "never"

155sandbox_mode = "read-only"

156```

157

158### Test the sandbox locally

159

160To see what happens when a command runs under the Codex sandbox, use these Codex CLI commands:

161

162```bash

163# macOS

164codex sandbox macos [--full-auto] [--log-denials] [COMMAND]...

165# Linux

166codex sandbox linux [--full-auto] [COMMAND]...

167```

168

169The `sandbox` command is also available as `codex debug`, and the platform helpers have aliases (for example `codex sandbox seatbelt` and `codex sandbox landlock`).

170

171## OS-level sandbox

172

173Codex enforces the sandbox differently depending on your OS:

174

175- **macOS** uses Seatbelt policies and runs commands using `sandbox-exec` with a profile (`-p`) that corresponds to the `--sandbox` mode you selected. When restricted read access enables platform defaults, Codex appends a curated macOS platform policy (instead of broadly allowing `/System`) to preserve common tool compatibility.

176- **Linux** uses `bwrap` plus `seccomp` by default.

177- **Windows** uses the Linux sandbox implementation when running in [Windows Subsystem for Linux 2 (WSL2)](https://developers.openai.com/codex/windows#windows-subsystem-for-linux). WSL1 was supported through Codex `0.114`; starting in `0.115`, the Linux sandbox moved to `bwrap`, so WSL1 is no longer supported. When running natively on Windows, Codex uses a [Windows sandbox](https://developers.openai.com/codex/windows#windows-sandbox) implementation.

178

179If you use the Codex IDE extension on Windows, it supports WSL2 directly. Set the following in your VS Code settings to keep the agent inside WSL2 whenever it's available:

180

181```json

182{

183 "chatgpt.runCodexInWindowsSubsystemForLinux": true

184}

185```

186

187This ensures the IDE extension inherits Linux sandbox semantics for commands, approvals, and filesystem access even when the host OS is Windows. Learn more in the [Windows setup guide](https://developers.openai.com/codex/windows).

188

189When running natively on Windows, configure the native sandbox mode in `config.toml`:

190

191```toml

192[windows]

193sandbox = "unelevated" # or "elevated"

194# sandbox_private_desktop = true # default; set false only for compatibility

195```

196

197See the [Windows setup guide](https://developers.openai.com/codex/windows#windows-sandbox) for details.

198

199When you run Linux in a containerized environment such as Docker, the sandbox may not work if the host or container configuration blocks the namespace, setuid `bwrap`, or `seccomp` operations that Codex needs.

200

201In that case, configure your Docker container to provide the isolation you need, then run `codex` with `--sandbox danger-full-access` (or the `--dangerously-bypass-approvals-and-sandbox` flag) inside the container.

202

203### Run Codex in Dev Containers

204

205If your host cannot run the Linux sandbox directly, or if your organization already standardizes on containerized development, run Codex with Dev Containers and let Docker provide the outer isolation boundary. This works with Visual Studio Code Dev Containers and compatible tools.

206

207Use the [Codex secure devcontainer example](https://github.com/openai/codex/tree/main/.devcontainer) as a reference implementation. The example installs Codex, common development tools, `bubblewrap`, and firewall-based outbound controls.

208

209Devcontainers provide substantial protection, but they do not prevent every

210 attack. If you run Codex with `--sandbox danger-full-access` or

211 `--dangerously-bypass-approvals-and-sandbox` inside the container, a malicious

212 project can exfiltrate anything available inside the devcontainer, including

213 Codex credentials. Use this pattern only with trusted repositories, and

214 monitor Codex activity as you would in any other elevated environment.

215

216The reference implementation includes:

217

218- an Ubuntu 24.04 base image with Codex and common development tools installed;

219- an allowlist-driven firewall profile for outbound access;

220- VS Code settings and extension recommendations for reopening the workspace in a container;

221- persistent mounts for command history and Codex configuration;

222- `bubblewrap`, so Codex can still use its Linux sandbox when the container grants the needed capabilities.

223

224To try it:

225

2261. Install Visual Studio Code and the [Dev Containers extension](https://marketplace.visualstudio.com/items?itemName=ms-vscode-remote.remote-containers).

2272. Copy the Codex example `.devcontainer` setup into your repository, or start from the Codex repository directly.

2283. In VS Code, run **Dev Containers: Open Folder in Container…** and select `.devcontainer/devcontainer.secure.json`.

2294. After the container starts, open a terminal and run `codex`.

230

231You can also start the container from the CLI:

232

233```bash

234devcontainer up --workspace-folder . --config .devcontainer/devcontainer.secure.json

235```

236

237The example has three main pieces:

238

239- `.devcontainer/devcontainer.secure.json` controls container settings, capabilities, mounts, environment variables, and VS Code extensions.

240- `.devcontainer/Dockerfile.secure` defines the Ubuntu-based image and installed tools.

241- `.devcontainer/init-firewall.sh` applies the outbound network policy.

242

243The reference firewall is intentionally a starting point. If you depend on domain allowlisting for isolation, implement DNS rebinding and DNS refresh protections that fit your environment, such as TTL-aware refreshes or a DNS-aware firewall.

244

245Inside the container, choose one of these modes:

246

247- Keep Codex's Linux sandbox enabled if the Dev Container profile grants the capabilities needed for `bwrap` to create the inner sandbox.

248- If the container is your intended security boundary, run Codex with `--sandbox danger-full-access` inside the container so Codex does not try to create a second sandbox layer.

249

250## Version control

251

252Codex works best with a version control workflow:

253

254- Work on a feature branch and keep `git status` clean before delegating. This keeps Codex patches easier to isolate and revert.

255- Prefer patch-based workflows (for example, `git diff`/`git apply`) over editing tracked files directly. Commit frequently so you can roll back in small increments.

256- Treat Codex suggestions like any other PR: run targeted verification, review diffs, and document decisions in commit messages for auditing.

257

258## Monitoring and telemetry

259

260Codex supports opt-in monitoring via OpenTelemetry (OTel) to help teams audit usage, investigate issues, and meet compliance requirements without weakening local security defaults. Telemetry is off by default; enable it explicitly in your configuration.

261

262### Overview

263

264- Codex turns off OTel export by default to keep local runs self-contained.

265- When enabled, Codex emits structured log events covering conversations, API requests, SSE/WebSocket stream activity, user prompts (redacted by default), tool approval decisions, and tool results.

266- Codex tags exported events with `service.name` (originator), CLI version, and an environment label to separate dev/staging/prod traffic.

267

268### Enable OTel (opt-in)

269

270Add an `[otel]` block to your Codex configuration (typically `~/.codex/config.toml`), choosing an exporter and whether to log prompt text.

271

272```toml

273[otel]

274environment = "staging" # dev | staging | prod

275exporter = "none" # none | otlp-http | otlp-grpc

276log_user_prompt = false # redact prompt text unless policy allows

277```

278

279- `exporter = "none"` leaves instrumentation active but doesn't send data anywhere.

280- To send events to your own collector, pick one of:

281

282```toml

283[otel]

284exporter = { otlp-http = {

285 endpoint = "https://otel.example.com/v1/logs",

286 protocol = "binary",

287 headers = { "x-otlp-api-key" = "${OTLP_TOKEN}" }

288}}

289```

290

291```toml

292[otel]

293exporter = { otlp-grpc = {

294 endpoint = "https://otel.example.com:4317",

295 headers = { "x-otlp-meta" = "abc123" }

296}}

297```

298

299Codex batches events and flushes them on shutdown. Codex exports only telemetry produced by its OTel module.

300

301### Event categories

302

303Representative event types include:

304

305- `codex.conversation_starts` (model, reasoning settings, sandbox/approval policy)

306- `codex.api_request` (attempt, status/success, duration, and error details)

307- `codex.sse_event` (stream event kind, success/failure, duration, plus token counts on `response.completed`)

308- `codex.websocket_request` and `codex.websocket_event` (request duration plus per-message kind/success/error)

309- `codex.user_prompt` (length; content redacted unless explicitly enabled)

310- `codex.tool_decision` (approved/denied, source: configuration vs. user)

311- `codex.tool_result` (duration, success, output snippet)

312

313Associated OTel metrics (counter plus duration histogram pairs) include `codex.api_request`, `codex.sse_event`, `codex.websocket.request`, `codex.websocket.event`, and `codex.tool.call` (with corresponding `.duration_ms` instruments).

314

315For the full event catalog and configuration reference, see the [Codex configuration documentation on GitHub](https://github.com/openai/codex/blob/main/docs/config.md#otel).

316

317### Security and privacy guidance

318

319- Keep `log_user_prompt = false` unless policy explicitly permits storing prompt contents. Prompts can include source code and sensitive data.

320- Route telemetry only to collectors you control; apply retention limits and access controls aligned with your compliance requirements.

321- Treat tool arguments and outputs as sensitive. Favor redaction at the collector or SIEM when possible.

322- Review local data retention settings (for example, `history.persistence` / `history.max_bytes`) if you don't want Codex to save session transcripts under `CODEX_HOME`. See [Advanced Config](https://developers.openai.com/codex/config-advanced#history-persistence) and [Configuration Reference](https://developers.openai.com/codex/config-reference).

323- If you run the CLI with network access turned off, OTel export can't reach your collector. To export, allow network access in `workspace-write` mode for the OTel endpoint, or export from Codex cloud with the collector domain on your approved list.

324- Review events periodically for approval/sandbox changes and unexpected tool executions.

325

326OTel is optional and designed to complement, not replace, the sandbox and approval protections described above.

327

328## Managed configuration

329

330Enterprise admins can configure Codex security settings for their workspace in [Managed configuration](https://developers.openai.com/codex/enterprise/managed-configuration). See that page for setup and policy details.

agent-approvals-security.md Codex Docs, 2026-03-05 06:22 UTC → 2026-04-23 12:28 UTC