Skip to content

Dline Privacy Policy

Effective date: October 25, 2026

Last updated: October 26, 2026

This Privacy Policy explains how the Dline maintainers process information through the Dline VS Code extension and Dline-operated services that the extension contacts. Dline is independently maintained community software.

The Dline maintainers, publishing under the Marketplace publisher identity tuvokyang, are responsible for the Dline-specific processing described in this policy.

Dline does not currently publish a dedicated private channel for privacy requests. Do not place personal data, API keys, access tokens, source code, or private prompts in a public Issue or in the vulnerability-reporting form. Suspected vulnerabilities must follow Report a Security Vulnerability.

Third-party model providers, MCP servers, websites, image services, identity providers, and other integrations are responsible for the information they receive under their own terms and privacy policies.

This policy covers:

  • the Dline VS Code extension;
  • Dline-operated documentation, support, and telemetry services under dline.cc;
  • diagnostic or support information you explicitly choose to provide to Dline.

This policy does not replace the terms of a model provider, MCP server, browser target, repository host, image provider, or other third-party service that you choose to use through Dline.

  • Dline stores its settings, credentials, task history, messages, activities, checkpoints, and recovery state locally by default.
  • Dline sends task content to a model provider only when a task makes a provider request using a Profile you configured.
  • External tools receive the request data needed for an operation only after you approve that operation or enable a matching auto-approval rule.
  • Usage reporting and error reporting are separate controls. Both are enabled by default in the current build, honor VS Code’s telemetry setting, and can be disabled independently.
  • Dline does not sell personal information or use task content for advertising.
  • Dline does not intentionally include prompts, source code, file contents, commands, terminal output, or user identity in ordinary remote usage or error-reporting payloads.

Dline may store the following information locally:

  • API Profiles, provider endpoints, model selections, and feature settings;
  • API keys, OAuth tokens, provider credentials, and MCP secrets in dedicated credential stores;
  • task messages, selected attachments, tool results, activities, usage records, and recovery snapshots;
  • workspace-scoped settings, capability toggles, rules, skills, workflows, hooks, and MCP descriptors;
  • command logs, diagnostic records, browser resources, image previews, and temporary input resources;
  • Git-based checkpoint metadata and task-history indexes;
  • pseudonymous installation or device identifiers used by enabled reporting channels.

Local information remains on your device unless a feature described below sends it to a configured service or you export and share it yourself. Local data remains until you delete the corresponding task, credential, setting, cache, diagnostic data, or Dline storage directory, subject to bounded cleanup performed by the extension.

Uninstalling the extension may not remove data stored outside VS Code’s extension installation directory. Use Dline’s deletion controls or remove the Dline data, home, and documents roots after backing up anything you need.

When you send a task or Dline continues an approved task, Dline sends the information needed to produce a response to the model provider selected by that task’s API Profile. Depending on the task and your choices, this can include:

  • your prompt and later feedback;
  • system instructions, enabled rules, skills, workflows, and tool definitions;
  • selected source code, files, images, documents, diagnostics, or workspace context;
  • previous conversation messages and model responses;
  • approved tool results, command output, browser results, MCP results, and generated summaries;
  • provider and model configuration required to route the request.

The provider determines how it processes, retains, and uses this information. Review the provider’s privacy policy and data-use settings before adding credentials. Dline cannot delete information already received by a third-party provider.

Dline can call external services only through capabilities available to the task and subject to the applicable approval policy.

An MCP server can receive tool arguments, resource identifiers, prompts, and authentication data required by that server. Workspace or global MCP configuration may be created by you or another project contributor. Review each server’s operator, source, permissions, and privacy terms before enabling it.

Web search, web fetch, and browser automation

Section titled “Web search, web fetch, and browser automation”

Search providers and fetched websites receive normal network information such as the requested query or URL, IP address, user agent, cookies already present in the selected browser context, and any form data or actions you approve. Browser automation can expose information displayed in or entered into the controlled page.

Commands execute on your machine and can contact any service named by the command. File tools can read or modify only within their admitted scope. Checkpoints use local Git repositories and may record file paths, hashes, and file contents in local checkpoint objects. Dline does not push checkpoint data to a remote repository unless you separately authorize a command or workflow that performs that remote write.

Image generation providers receive the prompt and any reference image or mask you approve. Model providers can receive images, PDFs, documents, or extracted text that you attach or that an approved tool returns.

Allow usage reporting is enabled by default in the current build. You can turn it off in Settings → About. VS Code’s telemetry.telemetryLevel and organization policy can further restrict delivery and cannot turn reporting on after you turn it off.

When enabled, Dline can send bounded operational counts and dimensions to otel.dline.cc, such as:

  • Dline version, release channel, operating-system and editor family;
  • provider, model, API format, and task purpose categories;
  • request outcomes, durations, retries, throughput, and token/cache/cost counts;
  • aggregate tool, MCP, task, subagent, and feature-use counts;
  • local drop counts and protocol health information.

Ordinary usage payloads are designed not to contain prompts, model responses, source code, file contents, file paths, commands, terminal output, task or workspace identity, account identity, API keys, or access tokens. Model and provider identifiers can still reveal information if you put confidential data into a custom identifier. Do not use account numbers, customer names, repository names, secrets, or other sensitive values as custom model or provider IDs.

Turning usage reporting off stops new collection for the remote usage channel, aborts delivery attempts, and deletes unsent usage records and the local usage access token as implemented by the current client. It does not delete records already accepted by the service.

See Telemetry and Issue Diagnostics for the current wire fields, limits, local queues, retries, and deletion behavior.

Allow error reporting is enabled by default in the current build and can be turned off independently in Settings → About. Host and organization telemetry settings can further restrict delivery.

When enabled, Dline can send allowlisted failure and runtime-health records to otel.dline.cc, including event type, severity, safe error code or fingerprint, provider/model classification, phase, retryability, duration, and bounded performance measurements.

Ordinary remote error-reporting payloads are designed not to contain free-form error messages, stack traces, prompts, model responses, source code, file contents, paths, URLs, commands, terminal output, task or workspace identity, account identity, API keys, or access tokens.

Turning error reporting off stops new remote diagnostics, aborts delivery, and deletes unsent diagnostic records and the local diagnostics access token as implemented by the current client. It does not delete records already accepted by the service.

Dline can create a diagnostic bundle only when you request an export. The bundle is written to your machine and is not uploaded automatically. Review it before sharing it.

A bug-report workflow can open a prefilled issue or authorization page after showing an approval preview. Information is not submitted to a public issue until you review and submit the destination form. Temporary higher-detail diagnostics, when available, require a separate issue-specific authorization and are not granted by ordinary usage or error-reporting consent.

Information you send in a support request is retained as part of that request and may be visible to repository maintainers or, for a public issue, to the public. Remove secrets and confidential project information before submitting.

Dline processes information to:

  • provide the extension, configured integrations, and requested support;
  • protect locally stored credentials and service access;
  • preserve task state and recover interrupted work;
  • enforce permissions, prevent abuse, and protect service security;
  • diagnose failures and improve compatibility, reliability, and feature design;
  • measure aggregate service use when reporting is enabled;
  • comply with legal obligations and respond to valid legal requests.

Depending on the applicable law and feature, processing is based on your request to provide the service, your consent or settings choice, Dline’s legitimate interest in security and reliability, or a legal obligation. You can withdraw optional reporting consent at any time through Dline settings.

Dline may disclose information to:

  • model providers and integration operators you configure;
  • infrastructure, hosting, authentication, payment, email, logging, and security providers used to operate Dline services;
  • GitHub, which processes repository activity and private vulnerability reports under its own terms and privacy statement;
  • maintainers or support personnel who need access to resolve a request or protect the service;
  • authorities or other parties when required by law or necessary to protect users, Dline, or the public;
  • a successor if responsibility for a Dline service is transferred, subject to applicable notice and law.

Dline does not sell personal information. Dline does not share task content with data brokers or use it for third-party advertising.

Dline follows data minimization and keeps information only for the period needed for the stated purpose:

  • local task, configuration, credential, and cache data remains until you remove it or a documented local cleanup limit applies;
  • unsent usage and diagnostics data is bounded by the client and is deleted when the corresponding reporting control is turned off;
  • accepted remote usage and diagnostic records are retained only for service operation, security, aggregate analysis, and troubleshooting, then deleted or irreversibly aggregated under the applicable service retention schedule when it is published;
  • support records follow the retention controls of the service through which you submit them;
  • private vulnerability reports are stored by GitHub and remain subject to GitHub’s retention and deletion capabilities as well as maintainer handling.

Each Dline-operated remote service is expected to publish its concrete server-side retention schedule. As of the last-updated date above, the public service documentation does not state a maximum retention period for accepted usage or error-reporting records. Because both reporting settings are enabled by default in the current build, turn off the corresponding setting in Settings → About if you do not accept an unspecified server-side retention period.

Dline uses scoped storage owners, owner-only credential files where supported, atomic local writes, bounded queues, encrypted transport for Dline-operated remote services, cancellation, and privacy allowlists. No system can guarantee absolute security.

Protect your device, VS Code profile, provider accounts, repositories, MCP configuration, and API credentials. Report a suspected vulnerability privately as described in Report a Security Vulnerability.

You can:

  • disable usage reporting and error reporting independently;
  • reduce or disable VS Code telemetry;
  • remove provider credentials, Profiles, MCP servers, tasks, attachments, and local Dline data;
  • decline a tool request or remove an auto-approval rule;
  • choose a local model or local integration where available;
  • review a diagnostic bundle or issue form before sharing it;
  • request access, correction, deletion, restriction, objection, portability, or withdrawal of consent where applicable law provides that right.

Dline does not currently publish a dedicated private intake channel for privacy requests. Do not send personal data through a public Issue or the vulnerability-reporting form. Until a private channel is published, use the controls above to stop optional reporting and remove local Dline data. Dline cannot fulfill requests for data controlled only by a third-party provider; contact that provider directly.

Dline maintainers and service providers may process information in countries other than yours. Where required, Dline will use an applicable transfer mechanism and contractual or technical safeguards. Your chosen model provider or integration can use different locations and safeguards under its own policy.

Dline is a developer tool and is not directed to children. Do not use Dline-operated services if you are below the minimum age required by applicable law or the relevant service terms.

Material changes will update the date at the top of this page and will be described in Dline release notes or another appropriate notice. A change to this policy does not silently enable a reporting control you disabled.

Privacy requests: no dedicated private channel is currently published; do not submit personal data through a public Issue or the vulnerability-reporting form

Security vulnerabilities: Report a Security Vulnerability

Repository and non-sensitive product issues: TuvokYang/dline