Privacy Policy
Effective date: October 25, 2026
Open Headers is committed to protecting your privacy and ensuring transparency about how our software operates. This Privacy Policy explains our data practices for the Open Headers browser extension and companion app.
1. Overview
Open Headers consists of:
- A browser extension for managing HTTP headers (available for Chrome, Firefox, Edge, and Safari)
- An optional companion desktop application for accessing local system resources
Every network call the software can make is publicly documented byte-for-byte in our wire-transparency documentation, so our privacy claims can be verified from the outside — with browser DevTools or a system-level packet capture — without trusting us.
2. Information collection and use
What we collect
We do not collect:
- Personal information or account identity
- Browsing history
- URLs, hostnames, header names, or header values
- Request or response data
- Rule contents, configurations, or any data from your computer
The only data that ever reaches us is the anonymous usage counting and the crash and error reports described in section 3 — which features get used and what kind of failure happened where in our own code, never what you use them on.
Local storage only:
- Your header configurations are stored locally in your browser using the browser’s storage API
- Dynamic header sources accessed through the companion app remain on your local device
- Your configurations and data are never transmitted to our servers or third parties
How information is used
All configuration data is used solely for the functioning of the extension and remains on your device. We have no access to this information.
3. Anonymous usage counting and crash reports (telemetry)
The browser extension, the desktop app, and the command-line tool count which features get used, and report when the software crashes or hits an error it did not expect. This helps us understand what to improve and what to fix, and both channels are designed so that they cannot carry your data:
- A fixed event vocabulary, compiled into the app. Events come in five
families — lifecycle markers (first run, session start, uninstall), milestones (the first
time an install does something: its first rule, first match, first request, first import,
first workflow run, first share sent or accepted, first server connection, first AI client,
first license), actions (a rule created or matched, a request sent, an import or a workflow
run, a share sent or accepted, a feature used, a surface opened, an upgrade prompt shown or
clicked, a limit reached, a license activated, an AI client connected), health (a typed
error code), and two anonymous website rows (an uninstall reason, a download click) — each
shape documented field-by-field in the
wire-transparency specification. Alongside its events,
every batch states how far the install has got — installed, configured, activated,
retained, team — and its scale, plan, role, team size, profile, active days and the tool it
migrated from, each a bucket or a word from a fixed list (“2–5 rules”,
“6–10 seats”, a team plan, an editor, an API client, “4–9 of
the last 28 days”), never what it did it with. Every field comes from a
fixed list — feature names, platform (such as
mac/win/linux), CPU architecture (such asx64/arm64), operating-system and browser major version (such as “macOS 15”, “Chrome 130” — never a build number), browser, app version, interface language, distribution channel (which store the app was installed from), how the install keeps its workspaces (local,server, orgit— never where), and coarse size buckets (such as “2–5 rules” or a window “at least 800 pixels wide” — never exact counts or sizes). Free-form text is structurally impossible, so URLs, headers, request or response data, rule contents, and file paths cannot appear in an event. - Anonymous, with an identifier you control. Events carry a random session id minted when the app starts (held in memory only, never saved) and a random install identifier that groups counts from the same installation — it is generated by a coin flip, never derived from your device, network, or any account, so it identifies the install, not you. Turning usage counting off (Settings → Application → About → Privacy) deletes it — a later re-enable starts with a fresh identifier that cannot be linked to the old one, so the switch doubles as a reset. There is no account, no fingerprinting, and IP addresses are not stored; our server records only the coarse two-letter country code its infrastructure resolves at the network edge, plus how old the install is in broad buckets (days “2–7”, never exact). Said plainly: no new identifier has been added since, but the plan, role, team size, sync kind, country and language stamped on every batch can make a small paying team in a small market distinctive. The licensing side, which knows who the paying teams are, is never joined to telemetry — the two share no identifier, payload or deployment — and the switch deletes the identifier, so nothing further can be filed under it.
- Crash and error reports carry no message and no identifier. When the software crashes or hits an error it did not expect, it sends what kind of error it was and where in the app’s own code it happened — never the message, never your data. A report carries only error class names, template ids (which standard engine message the error matched, never its text), and positions in Open Headers’ own files; nothing from a request, a rule, a script, or a path on your machine can be expressed, because every string on this channel must match a fixed grammar that cannot spell a URL, a path, a header, or a quote. Reports carry no install identifier at all: the random per-launch session id is all that groups them, so a report can never be joined to an install. No memory snapshot is ever taken, and log lines never travel.
- Inspectable, byte for byte. Settings → Application → About → Privacy → “View events” and “View reports” show every event and every report of your current session exactly as sent (or as suppressed, when a channel is off). The exact wire payloads are also published in our wire-transparency documentation.
- Disclosed, and off with one switch each. Both channels are disclosed here,
in the store listings, and in the settings themselves. The extension and desktop app have
one toggle per channel in Settings → Application → About → Privacy — “Anonymous usage counting”
and “Crash and error reports”, each independent of the other; the command-line
tool honors
OH_TELEMETRY=0(which silences both) or thetelemetryanderrorReportsconfig keys. Off means off — the channel goes completely silent, the install identifier and any counts not yet sent are deleted when usage counting goes off, and every feature keeps working. Turning crash reports off stops new reports from being recorded; a report recorded while the switch was on but not yet delivered stays on your device and is sent only if you turn the switch back on during the same session — otherwise it is discarded when the app closes. The command-line tool, whose process ends with each command, keeps a count or report it could not deliver in time in its configuration directory and sends it at the start of the next command; turning that channel off deletes what it kept. When the extension is uninstalled, the browser opens our site once so we can count the departure; that request carries only the random install identifier. - Hard-off on server surfaces. The self-hosted server, the web app it serves, and the MCP server never send telemetry and contain no code path to do so.
4. Data sharing and transfer
We do not share or sell any data. Anonymous usage counts and crash and error reports are received and stored in our own Cloudflare account; once a day the usage counts are copied into a database we run ourselves, where they are aggregated. Cloudflare is our infrastructure provider and the only data processor involved; no third-party analytics service or SDK is used, and nothing else is collected in the first place. Raw anonymous events are kept for thirteen months, then only aggregated counts remain.
5. Local communication
The browser extension and companion app communicate locally via WebSocket on port 8137. This connection:
- Is limited to localhost (127.0.0.1)
- Does not transmit data over the internet
- Requires explicit user action to enable
- Uses a simple JSON-based protocol for requesting and receiving header values
6. Permissions
The extension requires certain permissions to function:
- storage: To save your header configurations
- alarms: For scheduling header updates
- scripting: To initialize extension functionality
- declarativeNetRequest: To modify HTTP headers
- host_permissions (
<all_urls>): To modify headers for your specified domains
These permissions are used solely for the extension’s core header modification functionality and not for collecting information.
7. Security
We prioritize security through:
- Local-only data storage
- No external network connections beyond documented endpoints (license renewal, update checks, notification manifests, release notes, anonymous usage counting, crash and error reports) and destinations you configure yourself
- Publicly documented wire behavior that anyone can verify with a packet capture
- Regular security updates and a published vulnerability disclosure policy
8. Children’s privacy
Open Headers is a developer tool and not intended for use by children under 13 years of age.
9. Changes to this policy
We may update this Privacy Policy from time to time. We will notify users of any changes by updating the “Effective Date” at the top of this policy.
10. Transparency commitment
Open Headers is open-source software (Apache-2.0) with a verifiable privacy posture. We publish:
- A wire-transparency specification documenting the exact payload of every network call the software can make — including every telemetry event and every crash report
- A security whitepaper describing the architecture behind these guarantees
- A vulnerability disclosure policy
Because the software’s outbound behavior is fully documented, you can verify our privacy claims yourself — in the source code, or from the outside with browser DevTools or a system-level packet capture — no trust in us required.
11. Contact information
If you have questions about this Privacy Policy or the Open Headers project, please:
- Create an issue at github.com/OpenHeaders/open-headers
- Email us at contact [at] openheaders.com
We welcome feedback on both our software and our policies.
12. Consent
By using Open Headers, you consent to this Privacy Policy. As we do not collect any personal information, there is no personal data to manage or delete; anonymous usage counting and crash and error reports can each be switched off at any time in the app.
This Privacy Policy is provided to meet browser extension store requirements and to be transparent about our commitment to privacy and data security.