What AI coding tools send home in telemetry, and how to turn it off
What telemetry Claude Code, Codex, Gemini CLI, Copilot, Cursor, Cline, Continue, Aider and Zed collect, and the exact opt-out settings.
AI coding tools collect telemetry about how you use them. This audit covers nine popular tools: Claude Code, OpenAI Codex CLI, Gemini CLI, GitHub Copilot, Cursor, Cline, Continue, Aider and Zed. For each one it documents what the tool sends, where it goes, whether prompts or code are included, and exactly how to turn it off.
This research is based on official documentation and source code read at pinned commits. We did not run the tools or capture their network traffic. The key claims are listed with verbatim sources in the evidence ledger at the end.
Summary: what do AI coding tools send?
| Tool | On by default? | Data categories | Where it goes | Prompts/code in telemetry? | Opt-out mechanism | Evidence |
|---|---|---|---|---|---|---|
| Claude Code | Yes, with the Claude API. Error reports only for Pro, Max and Ultra sign-ins | Usage metrics, error reports | Anthropic, third-party logging infrastructure, a third-party error tracking service | Not in metrics. Error reports are redacted. /feedback sends your conversation, including code | DISABLE_TELEMETRY=1, DISABLE_ERROR_REPORTING=1, DISABLE_FEEDBACK_COMMAND=1 | Docs |
| Codex CLI | Yes for codex and codex exec. app-server is off unless launched with --analytics-default-enabled, as the VS Code extension does | Usage and health metrics: feature and config use, API and tool-call counts and durations | OpenAI (Statsig-format metrics to ab.chatgpt.com, analytics events to chatgpt.com). /feedback goes to Sentry | No, unless you opt in to otel.log_user_prompt for your own collector | [analytics] enabled = false in config.toml | Docs, Source |
| Gemini CLI | Usage statistics: yes. OpenTelemetry: no | Tool names, durations and success, model and API request metadata, session config | Google Clearcut (play.googleapis.com). OpenTelemetry goes to your own target | Usage statistics: no. OpenTelemetry: yes once enabled (logPrompts defaults to true) | privacy.usageStatisticsEnabled: false | Docs |
| GitHub Copilot | Yes | VS Code crash, error and usage telemetry. Copilot usage metrics. Server-side session data | Microsoft and GitHub | GitHub says Copilot CLI telemetry excludes prompts and code | telemetry.telemetryLevel: off (VS Code). --disable-client-telemetry (CLI, per run) | Docs |
| Cursor | No telemetry setting is documented | Privacy policy: usage, log and device data. Without Privacy Mode, code data may be used for training | Cursor (Anysphere) and its service providers | Training use, not telemetry, is what Privacy Mode controls | Privacy Mode covers training and retention only | Docs |
| Cline | Yes | Features, tasks, errors, token counts, timings, model and provider, version, OS, editor | PostHog via data.cline.bot | No, except for requests to the Cline and ClinePass providers | “Allow error and usage reporting” toggle, or telemetryOptOut | Docs |
| Continue | Removed in the final 2.0.0 release. Earlier versions: yes | Earlier versions: product analytics and error tracking | Earlier versions: PostHog (app.posthog.com) and Sentry | No | Update, or turn off “Allow Anonymous Telemetry” on old versions | Commit |
| Aider | No. It’s opt-in, and a random subset of users are asked | LLMs and token counts, edit formats, feature and command use, errors | PostHog (us.i.posthog.com) | No | --analytics-disable (permanent) | Docs |
| Zed | Yes | File extensions, features, project stats, frameworks, crash reports | Sentry, Snowflake, Hex, Amplitude | No | telemetry.diagnostics and telemetry.metrics | Docs |
Telemetry vs. model provider data retention
Client telemetry is what the tool itself collects about how you use it: features used, errors hit, session metadata. These are the settings this article covers.
Model provider data retention is separate. When you send a prompt to Claude, an OpenAI model or Gemini, the request goes to the provider’s servers. What they keep, and whether they train on it, depends on their policies and your account type, not on your telemetry settings.
- Anthropic: Under commercial terms (Team, Enterprise, API), Anthropic says it does not train on code or prompts sent to Claude Code, and standard retention is 30 days. Consumer accounts (Free, Pro, Max, Ultra) can allow training. With that setting on, data is retained for 5 years. Claude Code data usage
- OpenAI: Data sent to the OpenAI API is not used for training unless you opt in. By default, abuse-monitoring logs are kept for up to 30 days. OpenAI data controls
- Google: On paid Gemini API services, Google doesn’t use your prompts or responses to improve its products. On unpaid services, such as Google AI Studio and the free API quota, Google may use them, and human reviewers may read them. Gemini API terms
- Cursor: With Privacy Mode on, Cursor says it has zero-data-retention agreements with its model providers. Without it, Cursor may use and store your code data for training.
Turning off client telemetry stops the coding tool from collecting usage analytics. It doesn’t stop your prompts from reaching the model provider.
Detailed findings by tool
Claude Code
Default: Metrics are on with the Claude API. They’re off on Amazon Bedrock, Google Cloud’s Agent Platform (Vertex), Microsoft Foundry and Claude Platform on AWS, unless a host platform sets CLAUDE_CODE_PROVIDER_MANAGED_BY_HOST.
What is collected:
- Metrics: latency, reliability and usage patterns, “sent to Anthropic and to third-party logging infrastructure over TLS. Metrics never include your code, prompts, or file paths.”
- Error reports: error messages and stack traces from Claude Code’s own internals, sent to a third-party error tracking service. “Claude Code redacts known patterns of secrets, file paths, email addresses, and other personal information before anything leaves your machine.”
Error reporting is on only when all of these apply: you sign in with a Claude Pro, Max or Ultra subscription, you’re running v2.1.198 or later, you connect directly to the Claude API, and your organization doesn’t have a zero data retention or HIPAA agreement.
Feedback and surveys: /feedback is not telemetry, but it uploads the most. “When you run the /feedback command, a copy of your conversation history including code is sent to Anthropic.” /bug and /share go through the same path. Those transcripts are retained for 5 years. The “How is Claude doing this session?” survey records only your rating. If you then agree to share the transcript, it’s uploaded with known API key and token patterns redacted, but “source code, file contents, and other conversation content are uploaded as-is.”
Opt-out:
DISABLE_TELEMETRY=1: opts out of metrics. Any non-empty value opts out, even0orfalse. It also turns off feature-flag fetching, which can make Remote Control unavailable.DISABLE_ERROR_REPORTING=1: opts out of error reports.DISABLE_FEEDBACK_COMMAND=1: disables/feedback,/bugand/share.CLAUDE_CODE_DISABLE_FEEDBACK_SURVEY=1: disables the session-quality survey.DO_NOT_TRACK=1: has the same effect asDISABLE_TELEMETRY. Unlike the others, it’s read as a normal boolean, so0leaves telemetry on.CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1: turns off all of the above plus auto-updates, release notes, PR status badge checks and availability checks. If you set it, you have to update Claude Code yourself. It doesn’t affect the WebFetch domain safety check, which runs regardless of provider.
Source: Claude Code Data Usage and Environment Variables, accessed November 12, 2026.
OpenAI Codex CLI
Default: On for the CLI. OpenAI’s docs say: “By default, Codex periodically sends a small amount of anonymous usage and health data back to OpenAI.” In the source at commit c0cc9fc, the interactive TUI passes default_analytics_enabled = true (startup_orchestration.rs). codex exec sets DEFAULT_ANALYTICS_ENABLED: bool = true (exec/src/lib.rs). The exception is codex app-server, the mode other apps embed. PR #9118 says: “Analytics are disabled by default for app-server,” but first-party clients like the VS Code extension launch it with --analytics-default-enabled. When OpenAI announced the feature in Discussion #8291, it said analytics “will default to enabled, except in jurisdictions where opt-out is required by law.”
What is collected: OpenAI’s metrics catalog lists counters and timings, for example API request counts by status, WebSocket events, and tool calls with duration by tool name and outcome. Each one is tagged with auth mode, model and Codex version. OpenAI says “these metrics don’t contain any personally identifiable information (PII)”. The tool field names the internal tool, such as shell or apply_patch, not the command itself.
Where it goes: The default metrics exporter is Statsig (types.rs). It posts OTLP metrics to https://ab.chatgpt.com/otlp/v1/metrics (otel/src/config.rs). Analytics events go to …/backend-api/codex/analytics-events/events on chatgpt.com (analytics/src/client.rs). The separate, user-initiated /feedback upload goes to Sentry (feedback/src/lib.rs). Turn it off with [feedback] enabled = false.
Opt-out:
Add to ~/.codex/config.toml:
[analytics]enabled = falseThe source comment says: “When enabled is false, disables OpenAI analytics across Codex product surfaces on this machine.” A custom OTLP exporter you configured yourself keeps running. To stop that too, set otel.metrics_exporter = "none". Raw prompts only appear in OpenTelemetry logs if you opt in with otel.log_user_prompt.
Source: config_toml.rs, Advanced configuration: Metrics, Config Reference, accessed November 12, 2026.
Gemini CLI
Gemini CLI has two separate systems, and most of the confusion comes from mixing them up.
1. Usage statistics: on by default. The configuration reference lists privacy.usageStatisticsEnabled with “Default: true”, and the settings schema agrees. While it’s on, the Clearcut logger sends events to https://play.googleapis.com/log (clearcut-logger.ts). Google documents what it collects:
- Tool calls: the names of the tools called, whether they succeed, and how long they take. It says, “We do not collect the arguments passed to the tools or any data returned by them.”
- API requests: the model used, duration and success. It says, “We do not collect the content of the prompts or responses.”
- Session information: CLI configuration, such as enabled tools and approval mode.
While usage statistics are on, Gemini CLI also sends an installation ID header (x-gemini-api-privileged-user-id) with its model requests (contentGenerator.ts).
To opt out, add this to ~/.gemini/settings.json (it takes effect after a restart):
{ "privacy": { "usageStatisticsEnabled": false }}2. OpenTelemetry: off by default. The telemetry docs list enabled (GEMINI_TELEMETRY_ENABLED) with default false. This exports traces, metrics and logs to a target you choose: local (the default) or gcp, with a configurable OTLP endpoint. Note that logPrompts (GEMINI_TELEMETRY_LOG_PROMPTS) defaults to true. Once you turn OpenTelemetry on, prompts go to your collector unless you set it to false. To be sure it stays off:
export GEMINI_TELEMETRY_ENABLED=falseSource: docs/reference/configuration.md, docs/cli/telemetry.md, at commit aeb6c04, accessed November 12, 2026.
GitHub Copilot
Copilot data has three layers: VS Code’s own telemetry, Copilot’s client telemetry, and what GitHub collects on its servers.
VS Code extension
Default: On. The VS Code setup guide states: “Telemetry in your free version of GitHub Copilot is currently enabled.” It adds that you can opt out “by disabling telemetry in VS Code by setting telemetry.telemetryLevel to off, or you can adjust both telemetry and code suggestion settings in Copilot Settings.”
Opt-out:
"telemetry.telemetryLevel": "off"In VS Code itself, off stops crash reports, error telemetry and usage data. But the VS Code telemetry page warns that extensions “may be collecting their own usage data and are not controlled by the telemetry.telemetryLevel setting.” For Copilot, GitHub’s setup guide says this setting is the opt-out. Check every other extension separately.
GitHub-side data: Your local settings don’t change what GitHub records on its side. GitHub’s usage metrics docs say: “Usage metrics combine telemetry from IDEs, Copilot CLI, and other supported surfaces,” and “Server-side telemetry can still identify active users when client-side detail is missing.”
Copilot CLI
Default: Client telemetry is on. GitHub documents only a per-run opt-out:
copilot --disable-client-telemetryAccording to the CLI configuration guide, this “disables first-party telemetry sent by the CLI, including unrestricted and restricted telemetry. It is not saved in your configuration. The option does not disable GitHub’s server-side collection of session data or telemetry.”
Offline mode: The Copilot agents documentation says that when you use your own model provider, the CLI still sends telemetry to GitHub. That telemetry “does not include your prompts or code, but it does include usage metadata.” Setting COPILOT_OFFLINE=true turns off all telemetry, but only works with your own provider.
OpenTelemetry export (opt-in): Copilot CLI can export OpenTelemetry to your own collector. The CLI reference says: “OTel is off by default.” It starts when you set COPILOT_OTEL_ENABLED=true or an OTLP endpoint.
Source: VS Code: Set up Copilot, VS Code Telemetry, Configuring Copilot CLI, CLI command reference, Copilot agents, accessed November 12, 2026.
Cursor
Default: Cursor documents no setting that turns telemetry off. Its privacy policy describes collecting log information (“error logs, and other ways that you interact with the Service”) and usage data. It also names “analytics” among the service providers it shares data with.
What Privacy Mode controls: Privacy Mode is about training and retention. Cursor’s Data Use page says that with Privacy Mode on, “Customer Data will not be used for training by Cursor,” and that Cursor “maintains zero data retention (ZDR) agreements with all providers.” With it off, “we may use and store codebase data, prompts, editor actions, code snippets, and other code data and actions to improve our AI features and train our models.”
The only statement we found linking Privacy Mode to analytics is a January 2025 forum answer from a Cursor staff member: “you can enable privacy mode in your Cursor settings, which stops us from ever logging your queries for analytics.” That doesn’t appear in Cursor’s current documentation.
Also note: “Even if you use your API key, your requests will still go through our backend!” Cursor also temporarily caches file contents on its servers, encrypted.
Enable Privacy Mode:
- Open Cursor Settings (Cmd+Shift+J on macOS, Ctrl+Shift+J on Windows and Linux)
- Click General in the sidebar
- Toggle Privacy Mode on
According to Cursor’s help center, “For teams, Privacy Mode is enabled by default for all team members.” Admins can enforce it so members can’t turn it off.
Source: Cursor Data Use, Privacy and Data, Privacy Policy, accessed November 12, 2026.
Cline
Default: On. The docs say: “Telemetry is on by default in the VS Code and JetBrains extensions, Cline Desktop and the CLI, whether you sign in to a Cline account or use your own API keys.”
What is collected:
- Features used (tools, slash commands, skills)
- Tasks started and completed, and switches between Plan and Act
- Errors (API failures, tool errors)
- Token counts, timings, model and provider names
- Cline version, OS and editor
Each installation has a device identifier, and when you’re signed in, “events also identify your account and organization.” The docs say “Cline’s telemetry doesn’t include your code, file contents or prompts, except for requests to the Cline and ClinePass providers.” For those providers, “telemetry for these requests can include the prompt and the model’s response.”
Where it goes: PostHog, at the Cline-run host https://data.cline.bot (posthog-config.ts). The extension creates its client with new PostHog(posthogConfig.apiKey, { host: posthogConfig.host, ...}) (PostHogTelemetryProvider.ts).
Opt-out: One setting covers all Cline products on a computer.
VS Code and JetBrains:
- In the Cline panel, click the settings (gear) icon
- Select General
- Clear “Allow error and usage reporting”
CLI:
Set telemetryOptOut to true in ~/.cline/data/settings/global-settings.json. If you use --config <dir> or CLINE_DIR, the file is under <dir>/data/settings/ instead.
{ "telemetryOptOut": true}Caveats from the docs:
- In VS Code, “Cline also stops sending usage events while the editor’s own telemetry is off.” While the editor’s telemetry is on, Cline still sends a few required events, such as the one recording that you turned telemetry off.
- In Cline Desktop and the CLI, “other events continue until Cline and its background hub restart.” Run
cline hub stopor restart the computer. - With telemetry off, Cline still contacts its servers to sign you in, check feature flags and run requests on its own providers.
Reported issue: In Issue #7068 (October 2025, v3.34.0), a user captured requests to data.cline.bot after turning telemetry off. A Cline contributor said the service “emits several logs before realizing the telemetry is disabled.” The issue was closed as “not planned” in March 2026. We did not test whether current versions behave the same way.
Source: Cline Telemetry, Cline source at commits 8a6441f and 50e77bc, accessed November 12, 2026.
Continue
Default: Telemetry has been removed. In March 2026, the commit “remove all PostHog and Sentry telemetry collection” deleted core/util/posthog.ts, the Sentry logger and the posthog-node, posthog-js and @sentry/* dependencies. It kept only user-configured OpenTelemetry in the CLI. The README says the final 2.0.0 release included “removing anonymous telemetry”. It also says the repository “is no longer actively maintained and is read-only for all users.”
Earlier versions: Before 2.0.0, Continue sent PostHog analytics by default. The removed code created a client with new PostHog("phc_JS6X…", { host: "https://app.posthog.com" }) (posthog.ts before removal). The air-gapped guide tells older installs to “Turn off ‘Allow Anonymous Telemetry’ in the user settings. This will stop Continue from attempting requests to PostHog.” The config key is allowAnonymousTelemetry: false. The pre-removal CLI read CONTINUE_TELEMETRY_ENABLED=0 and CONTINUE_ALLOW_ANONYMOUS_TELEMETRY=0.
What to do: Update to the final release, or pin a version and turn the toggle off. The CLI’s remaining OpenTelemetry export runs only when you set an OTEL_EXPORTER_OTLP_* endpoint. CONTINUE_METRICS_ENABLED=0 or CONTINUE_CLI_ENABLE_TELEMETRY=0 disables it (telemetryService.ts).
Historical note: Issue #2082 reported telemetry sent to PostHog even with allowAnonymousTelemetry: false. A maintainer said they could not reproduce it. The issue was closed on March 19, 2026, the day the PostHog code was removed.
Source: Removal commit, README, Running Without Internet, accessed November 12, 2026.
Aider
Default: Off until you opt in. The analytics docs say: “Analytics are only collected if you agree and opt-in.”
There is a catch: “if you don’t provide --analytics or --no-analytics, aider will enable analytics for a random subset of users. Such randomly selected users will be asked if they wish to opt-in to analytics.” In the source, that subset is 10% of user IDs (PERCENT = 10 in analytics.py). The prompt’s default answer is Yes ((Y)es/(N)o [Yes]), so pressing Enter opts you in. “If you say ‘no’, analytics will be permanently disabled.”
What is collected:
- Which LLMs are used and with how many tokens
- Which edit formats are used
- How often features and commands are used
- Exceptions and errors
The docs say: “Aider respects your privacy and never collects your code, chat messages, keys or personal info.” Events carry a random UUID4 user identifier.
Where it goes: PostHog. The source at commit bdb4d9f contains:
posthog_project_api_key = "phc_99T7muzafUMMZX15H8XePbMSreEUzahHbtWjy3l5Qbv"posthog_host = "https://us.i.posthog.com"(analytics.py L56–57). The docs still say the default host is app.posthog.com, but the code uses us.i.posthog.com.
Opt-out:
Permanent:
aider --analytics-disableSingle session:
aider --no-analyticsConfig file (.aider.conf.yml):
analytics-disable: trueEnvironment variable: Aider reads every option from an AIDER_ variable (args.py):
AIDER_ANALYTICS_DISABLE=trueLog analytics locally without sending:
aider --analytics-log filename.jsonl --no-analyticsSource: Analytics, aider/analytics.py, accessed November 12, 2026.
Zed
Default: On. All Settings lists the default as "diagnostics": true, "metrics": true, matching default.json.
What is collected:
Client-side metrics:
- File extensions of opened files
- Features and tools used
- Project statistics (for example, number of files)
- Frameworks detected
Diagnostics: crash reports, made up of “a minidump and debug metadata,” sent on the next launch after a crash.
The docs say: “This data does not include your code or sensitive project details.”
Server-side telemetry: This is collected when you use hosted AI or Collaboration features, and is “required for these features to function.” Zed collects metadata such as token usage for rate limiting and billing. It says it “does not store your prompts or code unless you explicitly share feedback or opt into Edit Prediction training data collection.”
Where it goes: Events are sent to Zed’s servers “every 5 minutes (or when 50 events accumulate).” From there they’re routed to Sentry (crash reports), Snowflake (data warehouse), and Hex and Amplitude (dashboards).
Opt-out:
Open Settings (Cmd+, or Ctrl+,), search “telemetry” and turn off both settings. Or add this to settings.json:
{ "telemetry": { "diagnostics": false, "metrics": false }}Audit your telemetry: Run zed: open telemetry log from the command palette to see what Zed has reported.
Linked identity: “If you’ve authenticated, this ID may be linked to your email so Zed can analyze patterns over time and reach out for feedback.”
Source: Telemetry in Zed, All Settings, telemetry_events.rs, accessed November 12, 2026.
Disable telemetry everywhere: a copy-paste reference
Claude Code
export DISABLE_TELEMETRY=1export DISABLE_ERROR_REPORTING=1export DISABLE_FEEDBACK_COMMAND=1# Or turn off all non-essential traffic in one go.# This also disables auto-updates, so you'll need to update manually:export CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1Codex CLI
Add to ~/.codex/config.toml:
[analytics]enabled = false
[feedback]enabled = falseGemini CLI
In ~/.gemini/settings.json (usage statistics are on by default):
{ "privacy": { "usageStatisticsEnabled": false }}OpenTelemetry is already off by default. To make sure it stays off:
export GEMINI_TELEMETRY_ENABLED=falseGitHub Copilot (VS Code)
In VS Code settings.json:
{ "telemetry.telemetryLevel": "off"}GitHub Copilot CLI
Per run (not saved to your configuration):
copilot --disable-client-telemetryCursor
There’s no documented telemetry switch. To stop training use and get zero data retention with providers:
- Open Cursor Settings (Cmd+Shift+J / Ctrl+Shift+J)
- Go to General
- Turn on Privacy Mode
Cline
In the Cline panel: Settings > General > clear “Allow error and usage reporting”
Or in ~/.cline/data/settings/global-settings.json:
{ "telemetryOptOut": true}Continue
Update to the final 2.0.0 release, which has no anonymous telemetry. On older versions, turn off “Allow Anonymous Telemetry” in the settings.
Aider
Run once to disable permanently:
aider --analytics-disableZed
Add to settings.json:
{ "telemetry": { "diagnostics": false, "metrics": false }}Where Shadow fits
Shadow replaces secrets with reversible fakes in AI request traffic before prompts leave your machine. It masks model traffic from Claude Code, Codex CLI, Gemini CLI, GitHub Copilot, Aider, Cline, Continue and other agents that call a model API Shadow supports. Cursor uses a binary protocol, so Shadow recognizes its traffic but can’t mask it.
What Shadow doesn’t do: Shadow works on model request traffic, not on the telemetry these tools send to their analytics services. It doesn’t block or change Statsig, Clearcut, PostHog or Sentry calls. To turn those off, use the settings above. Shadow keeps your secrets out of the prompts and code that reach model providers, which is a separate problem from product analytics.
Limitations of this research
- We read source code and documentation. We did not run the tools or capture network traffic to confirm actual behavior.
- For closed-source tools (Claude Code, Cursor, GitHub Copilot), we relied on official documentation. There may be telemetry they don’t document.
- Source findings are tied to the pinned commits. Newer versions may behave differently.
- We did not audit the analytics vendors’ own data retention policies.
- We did not confirm that the opt-outs stop every network request to telemetry endpoints. Cline’s issue #7068 shows why that matters.
- Enterprise and team configurations may override individual settings.
About this research
Written by the Shadow Research Team, based on the public source code of six tools and the official documentation of three closed-source ones. Shadow is a privacy layer for AI from Tiny Bubble Company. It is not affiliated with any of the tools examined in this article.
Disclaimer: This article describes publicly available source code and documentation as of November 12, 2026. It is not legal advice. Behavior may change between versions, so check the settings on the version you have installed before relying on these findings.
Evidence ledger
These are the key claims, each with a verbatim quote or code excerpt and its source.
| Claim | Source | Verbatim quote or code | Date checked |
|---|---|---|---|
| Claude Code metrics are on by default with the Claude API | code.claude.com/docs/en/data-usage | “Metrics: latency, reliability, and usage patterns, sent to Anthropic and to third-party logging infrastructure over TLS. Metrics never include your code, prompts, or file paths. Set DISABLE_TELEMETRY=1 to opt out.” Table: “Default on.” | 2026-11-12 |
| Claude Code error reporting conditions | code.claude.com/docs/en/data-usage | “Error reporting is on only when all of these apply: you sign in with a Claude Pro, Max, or Ultra subscription…” | 2026-11-12 |
Claude Code /feedback includes code | code.claude.com/docs/en/data-usage | “When you run the /feedback command, a copy of your conversation history including code is sent to Anthropic.” | 2026-11-12 |
| DISABLE_TELEMETRY: any non-empty value opts out | code.claude.com/docs/en/env-vars | “Set to any non-empty value, such as 1, to opt out of telemetry. Setting it to 0 or false still opts out…” | 2026-11-12 |
| NONESSENTIAL_TRAFFIC also disables auto-updates | code.claude.com/docs/en/env-vars | “to disable nonessential network traffic: auto-updates, telemetry, error reporting, the /feedback command, … release notes…” | 2026-11-12 |
| Codex sends usage data by default | developers.openai.com/codex/config-advanced | “By default, Codex periodically sends a small amount of anonymous usage and health data back to OpenAI.” | 2026-11-12 |
| Codex TUI defaults analytics on | startup_orchestration.rs#L702 | /*default_analytics_enabled*/ true, | 2026-11-12 |
| Codex app-server defaults analytics off | openai/codex PR #9118 | “Analytics are disabled by default for app-server. Users have to explicitly opt in via the analytics section in the config.toml file.” | 2026-11-12 |
Codex analytics.enabled | config_toml.rs#L537 | “When enabled is false, disables OpenAI analytics across Codex product surfaces on this machine.” | 2026-11-12 |
| Codex Statsig endpoint | otel/src/config.rs#L11 | STATSIG_OTLP_HTTP_ENDPOINT: &str = "https://ab.chatgpt.com/otlp/v1/metrics" | 2026-11-12 |
| Gemini CLI usage statistics default on | configuration.md | “privacy.usageStatisticsEnabled (boolean): Description: Enable collection of usage statistics. Default: true” | 2026-11-12 |
| Gemini CLI Clearcut endpoint | clearcut-logger.ts#L244 | const CLEARCUT_URL = 'https://play.googleapis.com/log?format=json&hasfast=true'; | 2026-11-12 |
| Gemini CLI OpenTelemetry default off, logPrompts default on | telemetry.md | Table: enabled / GEMINI_TELEMETRY_ENABLED … false; logPrompts / GEMINI_TELEMETRY_LOG_PROMPTS … true | 2026-11-12 |
| Copilot telemetry on, VS Code opt-out | code.visualstudio.com/docs/setup/copilot | “Telemetry in your free version of GitHub Copilot is currently enabled.” | 2026-11-12 |
| VS Code setting doesn’t govern extensions in general | code.visualstudio.com/docs/configure/telemetry | “These extensions may be collecting their own usage data and are not controlled by the telemetry.telemetryLevel setting.” | 2026-11-12 |
Copilot CLI --disable-client-telemetry | docs.github.com configure Copilot CLI | “The option does not disable GitHub’s server-side collection of session data or telemetry.” | 2026-11-12 |
| Copilot CLI offline mode | docs.github.com Copilot agents | “If you enable offline mode by setting the COPILOT_OFFLINE environment variable to true, all telemetry is disabled.” | 2026-11-12 |
| Cursor Privacy Mode is about training and retention | cursor.com/data-use | “Customer Data will not be used for training by Cursor. Cursor maintains zero data retention (ZDR) agreements with all providers…” | 2026-11-12 |
| Cursor staff on analytics logging (Jan 2025) | forum.cursor.com | “you can enable privacy mode in your Cursor settings, which stops us from ever logging your queries for analytics” | 2026-11-12 |
| Cline telemetry on by default | docs.cline.bot telemetry | “Telemetry is on by default in the VS Code and JetBrains extensions, Cline Desktop and the CLI” | 2026-11-12 |
| Cline PostHog host | posthog-config.ts#L43 | host: "https://data.cline.bot", | 2026-11-12 |
| Continue removed PostHog and Sentry | continuedev/continue commit 32639c5 | “chore: remove all PostHog and Sentry telemetry collection” | 2026-11-12 |
| Aider analytics opt-in | aider.chat/docs/more/analytics.html | “Analytics are only collected if you agree and opt-in.” | 2026-11-12 |
| Aider uses PostHog | analytics.py#L56-L57 | posthog_host = "https://us.i.posthog.com" | 2026-11-12 |
Aider --analytics-disable | aider.chat/docs/more/analytics.html | “You can opt out of analytics forever by running this command one time: aider --analytics-disable” | 2026-11-12 |
| Zed telemetry defaults to true | zed.dev/docs/reference/all-settings | "telemetry": { "diagnostics": true, "metrics": true, …} | 2026-11-12 |
| Zed vendors | zed.dev/docs/telemetry | “Sentry: Crash-monitoring service… Snowflake: Data warehouse… Hex: Dashboards and data exploration… Amplitude: Dashboards and data exploration” | 2026-11-12 |
Install Shadow Core freeAll research
Frequently asked questions
How do I disable Claude Code telemetry?
DISABLE_TELEMETRY=1 turns off usage metrics, along with feature-flag fetching and session surveys. DISABLE_ERROR_REPORTING=1 turns off error reports. DISABLE_FEEDBACK_COMMAND=1 disables /feedback, /bug and /share. CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=1 turns off all of that and more, including auto-updates and release notes. For these variables, any non-empty value opts out, even 0 or false. Metrics and error reports are off by default on Bedrock, Vertex (Google Cloud's Agent Platform), Microsoft Foundry and Claude Platform on AWS.
Does Codex CLI send analytics by default?
Yes. In OpenAI's source, the interactive codex command and codex exec default analytics to on. OpenAI's docs say Codex 'periodically sends a small amount of anonymous usage and health data back to OpenAI.' To opt out, add [analytics] enabled = false to ~/.codex/config.toml. That turns off the built-in Statsig metrics exporter and Codex's analytics events. A custom OTLP exporter stays on unless you also set otel.metrics_exporter = "none".
Does Gemini CLI collect telemetry by default?
Its usage statistics, yes. privacy.usageStatisticsEnabled defaults to true and sends anonymized events to Google (Clearcut). Set it to false in ~/.gemini/settings.json to opt out. The separate OpenTelemetry setting, telemetry.enabled (GEMINI_TELEMETRY_ENABLED), defaults to false. When you turn OpenTelemetry on, logPrompts defaults to true, so your prompts go to the collector you configure.
How do I disable GitHub Copilot telemetry?
In VS Code, set telemetry.telemetryLevel to off. GitHub's setup guide says this opts you out of Copilot telemetry, and you can also adjust telemetry in your Copilot settings on GitHub. In Copilot CLI, copilot --disable-client-telemetry turns off client telemetry for that one invocation. COPILOT_OFFLINE=true turns off all telemetry when you use your own model provider. GitHub still collects server-side session data and telemetry either way.
What does Cursor's Privacy Mode do?
Privacy Mode stops Cursor from training on your code. It also means Cursor's zero-data-retention agreements with model providers apply. It's not a telemetry switch: Cursor's privacy policy describes usage, log and device data collection, and Cursor documents no setting to turn that off. A Cursor staff member wrote on the forum in January 2025 that Privacy Mode also stops Cursor logging your queries for analytics. Turn it on in Cursor Settings (Cmd/Ctrl+Shift+J) under General. For teams it is on by default.
How do I disable Cline telemetry?
In the Cline panel, open settings (gear icon), select General and clear 'Allow error and usage reporting'. In the CLI, set "telemetryOptOut": true in ~/.cline/data/settings/global-settings.json. One setting covers the VS Code and JetBrains extensions, Cline Desktop and the CLI on each computer. In Desktop and the CLI, some events continue until the background hub restarts (cline hub stop).
How do I disable Continue telemetry?
Update to the final 2.0.0 release (VS Code) or a current CLI. Continue removed its PostHog and Sentry telemetry in March 2026, and its README says the final release removed anonymous telemetry. On older versions, turn off 'Allow Anonymous Telemetry' in the user settings or set allowAnonymousTelemetry: false. The repository is no longer maintained.
How do I disable Aider analytics?
Run aider --analytics-disable once to opt out permanently, or use --no-analytics for a single session. Aider is opt-in: unless you pass a flag, it picks a random subset of users and asks whether they want to opt in. If you answer no, analytics are permanently disabled.
How do I disable Zed telemetry?
Open Settings (Cmd/Ctrl+,), search for 'telemetry' and turn off both Diagnostics and Metrics. Or add "telemetry": {"diagnostics": false, "metrics": false} to your settings.json. Server-side telemetry still applies when you use Zed's hosted AI or Collaboration features.