skip to content
legal

Privacy Policy

Effective date: June 15, 2026 · Last updated: November 3, 2026

the short version

TerminalHire is a developer job-matching tool: a command-line interface (CLI) plus a Claude Code plugin, together with the website terminalhire.com. Matching runs on your own machine, and it never sends your profile or a fingerprint to our servers. The website is different: signing in, posting and claiming paid tasks, and the analytics described below all run on our servers, and this policy says what each of them collects.

1. Who we are

TerminalHire is operated by Staqs, Inc. (“staqs”, “we”, “us”), registered at 251 Little Falls Drive, Wilmington, DE 19808, USA. For privacy questions or to exercise your rights, contact us at privacy@staqs.io.

Data protection / privacy contact: No Data Protection Officer has been appointed; privacy inquiries are handled by Staqs, Inc. at privacy@staqs.io. . EU/EEA representative under Article 27 GDPR: No EU representative has been appointed at this time; EU/EEA users may contact privacy@staqs.io. . UK representative, if applicable: No UK representative has been appointed at this time; UK users may contact privacy@staqs.io.

2. What data we handle

We group everything into four buckets, in order of how protective each is of your privacy.

2a. Data that stays on your device (never transmitted unless you consent)

The following lives only in your local TerminalHire installation and is never sent to us unless you take a specific opt-in action described in bucket 2c:

  • •An encrypted local profile (~/.terminalhire/profile.enc, encrypted at rest with AES-256-GCM): your skill tags, GitHub-derived public data, an optional display name and contact email you set yourself, and your saved jobs.
  • •A GitHub OAuth token, stored encrypted on your device, with scope read:user (public data only — we never request or access private repositories).
  • •Your profile never leaves your device — job matching runs against an anonymously downloaded job index. No fingerprint and no profile is sent to perform matching.
  • •Employer-repository sessions are excluded by default from profile building, to avoid capturing an employer's confidential or proprietary information.

2b. Anonymous processing (no developer identity)

  • •Job index download (GET /api/index): an anonymous download of the public job index. No cookies, no identifiers, and no profile are included in the request.
  • •Click tracking: links of the form terminalhire.com/j/<source>/<id> redirect (HTTP 302) to the real job listing and log an anonymous event consisting of the job id, source, company, and timestamp. This event contains no developer identity and no profile. Our hosting provider may transiently process IP addresses for security and operational purposes.
  • •Aggregate “matched” count: when the developer-directory match runs against a profile that never leaves your device, the CLI may report an anonymoussignal consisting only of the public GitHub logins of the developers and projects you were matched with, so each of those builders can see an aggregate “you were matched N times” count on their own dashboard. This request carries no cookie, no identifier, and no developer identity — it never includes your profile, your skill fingerprint, or who did the matching, and no per-viewer record is stored.
  • •Product analytics: the website records named steps on the two paths through it — posting a paid task, and claiming it — so we can see where people get stuck. Each event carries a step name, a step number, and a few tags (which refusal a publish hit, whether a posting is paid), plus which of our deployments recorded it and — on product events — which of our own surfaces it came from (web, CLI, MCP or our verification worker); error reports carry the deployment but not always the surface. Both describe our software, not you. While you are signed in it is tied to your public GitHub login; signed out, each event gets a throwaway id and no profile is built. We never send an email address, a token, a session id, repository contents, a patch, or the text of a posting. There is no autocapture — we record the steps we named, not every click you make. We do record a masked session recording: the shape of the page and what you clicked, with all text and every form field masked in your browser before anything is sent, so a recording shows where you moved and got stuck, and the address of each page with any share-link token or checkout session id taken out, but not the words on it or anything you typed. We also send error reports and server logs when something breaks. An error report carries a stack trace and the error message, with addresses and token-shaped strings redacted from the message. The homepage also sets one cookie, th_hp, holding only the letter a or b, so you keep seeing the same version of the page for 30 days; that letter is sent with the posting steps recorded on the homepage, on /auditand on a project's audit page; nothing else is stored under it, and it identifies nobody. No advertising or cross-site trackers, and no analytics cookies beyond that one letter. See the Cookies & analytics section below.

2c. Consented data egress (explicit, per-event, opt-in)

These features send identifying data when you choose to use them. The email signup below asks you to confirm your address before we add it to the confirmed list.

Email updates about paid tasks

If you leave your email, we store your email address, an optional plain-text note of up to 280 characters, an optional link to your LinkedIn profile, and an optional campaign tag. We keep only the profile address (https://www.linkedin.com/in/<name>) and drop anything else in the link you paste. We never fetch or read your LinkedIn page. When your signed-in session contains your numeric GitHub id, we also store that id and your GitHub login. Without that id, we store neither. In this signup flow, the login is for display; only the numeric id links the entry to your account. We use these details to write when paid tasks open up. Our admins read them on an internal page, and confirming sends Eric an email with your address, login, note, LinkedIn link and campaign tag. Confirming does not reserve a place or grant access to paid tasks.

terminalhire waitlist join makes the same signup from the terminal. It sends the address you give with --email, the note and LinkedIn link if you give them, and the campaign tag cli-waitlist. The CLI checks each field on your machine first and sends nothing if one fails. Of a pasted LinkedIn link, only the profile address leaves your machine. If you have linked this terminal to your account with terminalhire link, the command first sends your session cookie to ask us which GitHub login that session carries, so it can show you that login. The signup then goes out with the same cookie, so the rule above about your numeric GitHub id applies. An unlinked terminal sends no cookie, and its entry carries no login or id. The command never sends your GitHub token, your profile or your fingerprint.

We send a confirmation email before adding you to the confirmed list. Submitting the signup form again cannot replace the details on a confirmed entry. If an unconfirmed entry has not had a confirmation send attempt in the past hour, a new submission replaces its note, LinkedIn link, login, numeric id and campaign tag. Unconfirmed entries are deleted after 7 days. The signup row stores no IP address or user-agent. The list lives in our database, and Resend delivers the emails.

Signup and confirmation do not fetch your GitHub repositories or following list. Older operator tools remain available to read public repository languages and your public following list to select open-source issues; they are not run by this signup flow, and signup promises no weekly issue.

  • •Tier-1 opt-in profile sync. When you run terminalhire sync --pushand type “yes” to a consent card that names “staqs (terminalhire.com)”, a one-time snapshot of your GitHub-public profile fields (GitHub login, name, public email, top languages, skill tags) plus the display name and contact email you set yourself is stored on our server (a Neon PostgreSQL database). It is revocable at any time with terminalhire sync --delete, which performs a hard delete. This sync never stores private repositories, employer-repo-derived tags, raw code, access tokens, session context, or file paths.
  • •Opt-in dashboard claim view. When you run terminalhire claim --push, type “yes” to a field-level consent card, and confirm in the browser (signing in with GitHub to prove the account is yours), a minimal, score-free snapshot of your own posting and contribution claims — the kind of task, the public repository, a coarse status (claimed / in-progress / merged), the public pull-request link, a merged yes/no flag, and timestamps — is stored on our server (a Neon PostgreSQL database) so it can appear on your own dashboard. It never includes your diff-acceptance score, review notes, branch names, worktree paths, repository policy results, or any private data. You can hard-delete it at any time with terminalhire claim --push --revoke or the “delete my pushed claims” button on your dashboard.
  • •Lead sharing. Per opportunity, you are shown an explicit named-buyer prompt (for example, “Share with Northstar Talent Partners? yes/no”), and only if you answer “yes” is a lead payload (the approved profile fields) sent to that named partner agency. Each partner agency signs a per-partner DPA before it can receive any lead. The partner agency is an independent controller for recruiting purposes, governed by our Data Processing / Data Sharing Agreement.
  • •Submitting a claim. When you submit a claim, the pull request we open in the repository of the poster carries, in its body, the commit and branch, the task the poster asked for, the description you supplied with --body-file or PR-BODY.md, the paths your patch changed, the screenshot captions you supplied and links to those screenshots on TerminalHire (they open only for you and the poster, signed in), and the names of any workflow files our scan matched; we store that description with your run record, or with your change while it waits for the approval of the poster, and replace anything in it that looks like a credential; each @ in your words is replaced with a look-alike character so it notifies nobody on GitHub, and anyone who can see that repository can see the pull request.

2d. Code audits

When you order a code audit, a developer working for Staqs reads your repository to do it. Our developers may use AI tools to help them, only under settings that do not allow your code to be used to train AI models. We delete our copies of your code within 7 days of delivering the report, and keep the report.

When your order is paid, we email a confirmation to the address you gave at checkout, and our team gets a notice of the order that includes that address. Our email processor (Resend) delivers both. The answers you give on the quote page (which units to focus on, your question, context, what worries you) are stored with the quote and the order, sent to the reviewer with the order, and deleted with your account.

Your human audit report contains the review, findings, quoted code lines and codebase write-up. We keep the report after deleting our copies of the repository. A full refund hides the report from your audit page but does not delete it. We delete the report when you delete your account. To ask us to delete a report without deleting your account, email eric@staqs.io. Payment records are kept for tax and accounting. We use Resend to email your checkout address when the report is ready. That email names your repository and the reviewed commit and links to your audit page.

Behavior cards and test packs. A human audit may also store behavior cards, checked source excerpts, card revisions, your Right/Wrong/Not sure decisions, your policy text and explanations, and reviewer-created test patches and run notes. Cards, decisions and packs record the reviewed commit, author and time. Previous versions remain in the private audit record; a changed answer can make a pack unavailable to download. These records are visible to the owner and authorized audit reviewers. A full refund hides them, and account deletion deletes them together with the report.

Agent audit.If you buy an agent audit on the audit quote page, our AI agent reads your repository at the commit we quoted: every application-code file the quote counted. It never reads .env files, other secret or key files, or symlinks. We send those files to Anthropic's API, which runs the Claude Sonnet model under Anthropic's commercial terms; those terms do not allow your code to be used to train models. We do not keep the contents of the files the agent read. We keep the findings, including the lines they quote, and the paths of any files it could not read, until you delete your account. The agent audit runs only after you pay for it. An AI agent does it, not a person; the human audit is done by an expert developer.

For an audit quote request, we keep the repository, commit, detected areas and your self-reported booking preference with your account. When you delete your account, we delete any audit request you have not paid for. An order you paid for, or one whose payment page is still open, stays as a payment record, and we delete your booking preference from it. Saving a card stores it with Stripe, not with us, and does not pay for the audit; payment requires your later acceptance of a price. If you open the booking calendar, Google Calendar (part of Google Workspace) loads only after your click. When you book, Google receives your name, email and chosen time under Google’s own terms. We do not send Google your repository name, detected areas or credentials. A booking marked in TerminalHire is your own report, not a confirmation from Google. When you submit the request, our team receives your repository, commit, confirmed contact email and booking preference through Resend.

2e. Notification and reminder emails

Submission emails to the poster can include the repository pull-request link, branch and commit, and workflow file names from the secret-exposure scan. Resend processes this message content; we do not send workflow contents or secret values in it.

If you post a paid task, we email you when something on your posting needs you. The email about a note a developer wrote to you carries the note's words, quoted line by line, unless the note was sent as a question; a developer's note therefore passes through our email processor and rests in the poster's inbox, where our deletion clocks do not reach. If you claim one, we email you when the poster acts on your claim; a change request, a message from the poster and an answer to your question each carry the poster's own words, and every other claim email names the posting only by its short reference. A poster's message is emailed only while you hold the claim. We send to the contact email you confirmed with us, or else the public email on your GitHub profile, read when we send; with neither, no email goes out.

When something we emailed you about is still waiting on you a day later, a setup step you started (saving a card, confirming an email) is unfinished, or an intro request to you is unanswered, we may send up to three reminders, a day apart, to the address the first email used. A reminder about confirming an email is not sent to the address waiting to be confirmed, unless that address is also the public email on your GitHub profile, and apart from a task's title, no reminder carries anything a person typed. To send each reminder only once, we keep a record of which reminder it was, an id for the claim, posting, intro, draft or account it was about, and when it was sent. The record does not hold your address, and we delete it after 30 days. Records of reminders about saving a card or confirming an email are also deleted when you delete your account, and from then on no reminder is sent to your account. Every one of these emails has a one-click link to stop it, except the emails about a payment we are holding or have already taken, which always go out; and your settings have a switch for each kind, reminders included. Our email processor (Resend) delivers them.

For a task on a project, the emails to its poster name the task by its title, show its listed price, and link to the task's own page. The email about delivered work can also give the test counts from our run, how many screenshots the developer attached, and a link to the delivered changes in your own repository on GitHub, whose address holds the repository name and commit ids. When a developer's question on such a task has had no answer for a day, we send the poster one reminder, between 9am and 8pm New York time, listing the titles of the tasks waiting on them. Neither email includes the question or the developer's name. The reminder is recorded and stopped the same way as the reminders above.

When a new paid task appears on the public listing, we email every developer we have an address for, once per task, unless you have turned these emails off. That is every account with a contact email confirmed with us or a GitHub public email on a current sign-in, except the task's own poster, an account that has posted tasks and never claimed one (we treat it as a poster, not a developer, and do not send it these emails at all), an account set to poster-only, an account whose posting access we revoked, and a deleted account. We send it to your confirmed contact email, or else the public email on your GitHub profile, read when we send. It carries only what the public listing shows: the task's public title, its price, its short reference and a link to claim it, never the poster's name or the repository. Each one has a one-click link that stops these emails and only these, and your settings have a switch for them; turning off all claim emails or all posting emails does not turn these off. To send each one only once, we keep a record of the task, your numeric GitHub account id, when we sent it and whether it went. The record does not hold your address, and we delete it after 30 days, or when you delete your account, after which none is sent to your account. Separately we keep, with no end date, a list of every task we have come across and whether we announced it; it names no person.

2f. Project records and membership

We store your project record, delivery policy, field states and sources, revisions, and the repository/commit and exported text confirmed. We also store team proposals, notes, questions and membership start and end records. Each team entry records its author and time; proposals record acceptance or dispute, and questions record resolution. The owner and authorized membership operators can read these records. Audit-review access alone does not grant membership access. Authorized membership operators can also set up a project for you: they create the project record under your account, can add repositories your GitHub App installation already covers, and can start a membership. Your project page then says it was set up by the TerminalHire team.

On a project with a live membership, our team may record the project details for you: it accepts its own proposals into your record and confirms that version without asking you first, and the record shows that the team recorded it, with the operator’s GitHub login in our history of the record. We show a notice on your project page and, unless you have turned our emails off, send you an email. A developer sees the record only after you approve that developer. You can change the record at any time, and later tasks use the version you save with Save and use this version; a task already posted keeps the version it carries. Our team does not record over a value you wrote or accepted: its proposal waits for you. Without a live membership, a team proposal does not change your record until you accept it. Notes are observations, not instructions, and remain outside the exported record. Confirmed policy fields appear in PROJECT_CONTEXT.md and travel with the task context supplied to developers. Pending team proposals and team questions do not. Operators can request deterministic suggestions from repository configuration through your existing GitHub App installation; the stored suggestions name their source paths and commit. No model call is used to produce these suggestions. We also run this read ourselves, without an operator asking, when a repository is added to a project and when the default branch of the repository holding the project record changes. Each suggestion is saved as a proposal and does not change your record by itself. Operator-recorded memberships describe the service period. Paid memberships also have separate subscription, invoice and credit records, retained as described in section 7.

We also store each project’s repository list: which of your repositories belong to it and which one holds its record. The repository list routes tasks; it does not change what TerminalHire can read on GitHub, which your GitHub App installation alone decides. We store a services map that you, or a connector you allowed to act, write: services, environments and repositories we cannot reach, with their providers, purposes, addresses and who administers them. The services map stays outside the exported record and is never given to developers. You, a connector you allowed to act, and authorized membership operators can read all of it; a read-only connector reads kinds, names, providers, purposes and environments only. Each team view of the services map is logged with the operator’s login and the project. Removing an entry removes it from the live map; copies may remain in backups, in the proposal that suggested it, and in logs until those expire.

You can invite a GitHub account to one project as a viewer or collaborator. After accepting, they can see your project record, delivery policy, repository names, people list, proposals, notes, questions, task briefs and status, submitted work summaries, branch and pull request links, check results, screenshots and project feedback with attachments. Both roles see only service kinds, names, providers, purposes and environments, including in service proposals. They do not see credit, billing details, task prices, satisfaction ratings or legacy owner-private notes. Sharing gives them no access to your repository on GitHub. A collaborator can add project notes, questions, proposals and task feedback; a collaborator cannot approve a proposal or a held change, confirm the record, publish a task, change App access, save a card, spend credit or accept work.

We store the invited GitHub account’s numeric id and login, role, inviter, invitation expiry and acceptance or removal times, a hash of the invitation token, and events recording invitations, acceptance and removal. Invitation links expire after seven days and acceptance checks the signed-in numeric account id. The owner can see and remove access from the people list. Removing someone ends their access on subsequent requests; it does not delete what they already read or contributed. If an invited person deletes their account, their access ends and their account ids, stored login, invitation hash and event labels are cleared. Their entries stay with the author shown as “former collaborator” until the owner deletes the project or their account.

2g. Team connectors

A membership operator can create a team connector: a token their agent presents to our team server. It reaches one project, and only while that project’s membership is live. It can read that project’s record, proposals, notes, questions and membership and how far its tasks have got, and it can record proposals, notes and questions there under the operator’s GitHub login, shown with the connector’s name. It can also prepare a priced, unpublished draft of a task on that project’s repositories, which the owner reviews, changes and publishes themselves. It can also write a plain-language explanation of a change held for the owner’s approval. We store it with that change, together with the operator’s GitHub id. The owner sees it marked as written by the project team, and developers do not see it. An operator can also create a team connector for reminders, which can do one thing: take the reminders now due to the project’s owner, so that our team’s agent can send them in the group chat the owner uses with us, such as Signal. A reminder says that delivered work waits for the owner’s review, or that its review date is near. It carries the task’s title, its listed price, dates and a link to the owner’s page. It carries no code and does not name the developer. We store which reminders were handed out and when: the task, the delivered work, the project, the connector and the batch it went to, and no message text. We delete these records after 30 days. Reminders are sent for a project only after our team has set up a reminders connector for it, and none is handed out once you ask us to delete your account. A switch for the owner to turn them off is not built yet; until it is, tell us and we turn them off for your account. During the pilot our team may instead publish a team draft for a pilot member, and the operator’s GitHub login goes on the posting’s event log and, when the team took the whole-repository notice for them, on the record of that notice. If your project record is confirmed it goes on the task; if not, the operator has to choose to publish without it. An operator can also add the confirmed project record to an open task of yours, on a project with a live membership, when the task needs approval and has no claims. The record goes into the private text of the task, which a developer reads after you approve them, and the title and public description do not change. Each such change goes in our server logs with the GitHub login of the operator, the project and the task. A task with a project record says so on its public preview, without naming the project, repository or owner: the preview and our public index carry that yes and nothing from the record. For each draft we hold who prepared it (the operator’s GitHub id and the connector’s name), the proposed price and the reason given. We store a hash of the token, never the token itself, with the operator’s GitHub id and login, the name they gave it, the project, and when it was created, last used and revoked, and when it expires. It expires 30 days after it is created, and it works only while that login is on our operator list. It cannot start or end a membership, change a project record, accept or dispute a proposal, publish a task or move money. The operator can revoke it at any time from the project page or their dashboard. If the operator deletes their account, their team connectors are deleted; the proposals, notes and questions they wrote stay in the owner’s project record with the author shown as “former operator”, and are deleted when the owner deletes that project or their account. A draft they prepared keeps its text and price and shows “former operator” in place of who prepared it. An explanation of a held change keeps its text, and we remove the operator’s GitHub id from it.

2h. Agent approvals

Each of your projects has a setting, “Agent may approve”, which is off unless you turn it on in the project’s Overview. Only you can change it, and only signed in. While it is on, a connector you allowed to act, or your project team’s agent when you ask it to, can approve a developer’s held change for you. We store the setting with your GitHub id and when it last changed, and a log of each time you turned it on or off. Both are deleted when you delete the project or your account. When a change is approved we record how: by you on the task page, by your connector, or by the team’s agent, with which connector it was and, for the team’s agent, the GitHub login it gave as the person who asked, which must be yours. The login and the connector are cleared when you delete your account. Approving puts the change on its branch; accepting the work and paying still happen only in your browser.

3. Purposes and lawful bases

For developers in the EU/EEA and UK, we rely on the following lawful bases under the GDPR / UK GDPR:

  • •Consent (Art. 6(1)(a)): Tier-1 profile sync and each lead share. You give consent through the named, per-event prompts described above, and you may withdraw it at any time.
  • •Legitimate interests (Art. 6(1)(f)): anonymous click analytics, product analytics (the named posting and claiming steps described in section 4), and the security/operational processing of transient IP addresses — to understand aggregate engagement, to find where the product gets in your way, and to protect the service. We have weighed this against your rights and freedoms (balancing test); the click events carry no developer identity at all, and the product-analytics events carry your public GitHub login and a step name and nothing more, so the impact on you is minimal. You can object — see your rights below.

4. Cookies & analytics

terminalhire.com sets no advertising or cross-site cookies and runs no advertising or cross-site trackers; the one analytics-related cookie is the one-letter th_hp described in section 2. The TerminalHire CLI and plugin do not use cookies.

We collect two usage signals. The first is the first-party anonymous click event described in section 2b (job id, source, company, timestamp — no developer identity, no profile), which is logged server-side and sets no cookie. The second is product analytics, processed by PostHog (see section 5): named steps on the posting and claiming paths, so we can see where people get stuck. It stores a local-storage entry in your browser holding an identifier for these events, and no cookie. While you are signed in the events carry your public GitHub login; signed out, each gets a throwaway id and no profile is built. Autocapture is off, so what is recorded is the named steps and their tags — never an email address, a token, repository contents, a patch, or the text of a posting. When we read a poster's repository to scope a posting we also record how big it was— counts of files, bytes and lines, the repository's numeric GitHub id, and whether we downloaded it or reused a copy, or which size limit stopped us reading it — and never a file name, a path or any of its contents. The copy we read is the repository at the scoped commit, kept on one server's temporary disk and reused for at most an hour after it was downloaded, never after; an expired copy is deleted the next time that server downloads a repository, or when the server is recycled. Your GitHub App access is checked again before every read.

PostHog also receives masked session recordings, error reports and server logs. A recording captures the structure of the page and your interactions with it — clicks, scrolling, navigation — and every piece of text and every form field is replaced with a placeholder in your browser before the recording is sent. That means a recording can show us that you stopped on the payment step and clicked the same button three times; it cannot show us your email address, the text of your posting, or anything you typed. It does carry the address of each page, as the analytics events from your browser do, and some addresses name a public repository (/projects/owner/repo). Before anything is sent, the token is taken out of every share, claim or draft link (the part after /c/, /t/, /dashboard/work/ or /poster/confirm/). We record campaign source, medium, and campaign names from links you follow to understand how people find TerminalHire. Error reports carry the error type, a stack trace, and the message with addresses and token-shaped strings redacted before they are sent. Server logs are lines our own code writes about what it did — never request bodies.

Beyond th_hp, the only cookies we set are strictly necessary session cookies that keep you signed in to authenticated areas (the dashboard, recruiter portal, and admin) when you choose to log in. These are exempt from prior-consent requirements because they are essential to a feature you explicitly requested. Our hosting provider may transiently process IP addresses for security and operational purposes, as described above.

5. Sub-processors & third parties

PartyPurposeLocationData involved
Vercel Inc.Website hosting, edge/CDN delivery of terminalhire.comUnited StatesAnonymous web requests; transient IP for security/operations
Neon Inc.Managed PostgreSQL database storing consented Tier-1 profilesUnited StatesConsented Tier-1 profile fields (only if you push a sync)
GitHub Inc.OAuth identity provider (you authenticate directly with GitHub)United StatesOAuth scope read:user — public profile data only
Google LLCGoogle Tag Manager / analytics on the websiteUnited StatesWebsite analytics events and cookies
PostHog Inc.Product analytics, error reports, server logs and masked session recordings for the website — so we can see where people get stuck and why something brokeUnited StatesStep name and number, a small set of tags (which refusal a publish hit, whether a posting is paid), which of our surfaces and deployments produced the event, the address of each page you view with share-link tokens and checkout session ids taken out, and your public GitHub login while you are signed in. Error reports carry the error type, a stack trace, and the message with addresses and token-shaped strings redacted. Session recordings carry the SHAPE of a page — layout, clicks, scrolling — with all text and every form field masked before it leaves your browser. No email, no token, no repository contents, no posting text, no readable page text.
ResendDelivers the email our service sendsUnited StatesThe recipient email address and the content of the message
Stripe, Inc.Payments: card checkout and saved cards for posters, and payouts to developersUnited StatesThe payment and payout details you enter with Stripe, your email address, and order amounts
Anthropic, PBCRuns the agent audit on the repository files you send itUnited StatesEvery application-code file in your repository at the quoted commit, only when you buy an agent audit.
Each partner agency (per-partner DPA)Named recipient of consented leads (independent controller for recruiting); each partner agency signs a per-partner DPA before it can receive any leadUnited StatesOnly the approved lead fields you explicitly chose to share with that named partner agency

This list is current as of October 5, 2026. We will update it as our sub-processors change.

6. International transfers

Our servers and the third parties above are located in the United States (Vercel and Neon host in the US). If you are in the EU/EEA or the UK, any transfer of your personal data to the United States relies on appropriate safeguards — the EU Standard Contractual Clauses (SCCs), the UK International Data Transfer Agreement (IDTA) or Addendum, and/or an adequacy mechanism, as applicable. Specifics to confirm: [to be finalized with counsel: confirm transfer mechanism per recipient (SCCs / UK IDTA / adequacy / DPF certification)].

7. Data retention

The unsubscribe link on the confirmed page and in later updates deletes your list entry, including the note, LinkedIn link, login and numeric id, plus any stored issue picks, send history and public-following cache associated with it. Account deletion uses only your numeric GitHub id to find and delete linked entries and their associated records, even if your login has changed. It never matches entries by login. Entries without a numeric id stay on the email list and use their own unsubscribe link. Database deletion does not remove an interest notice already delivered to Eric's mailbox; contact us to request deletion of that copy.

  • •On-device data: kept locally on your machine until you delete it. You control it directly.
  • •Tier-1 synced profile: retained until you revoke it (terminalhire sync --delete) or after 24 months of inactivity, whichever comes first.
  • •Anonymous click logs: retained for 12 months , then deleted or further aggregated.
  • •Shared leads: once a lead is shared with a named buyer, that buyer becomes an independent controller and applies its own retention, governed by the per-partner DPA.
  • •Our record of each payout we made to you stays, with your GitHub account id on it, because we keep payment records for tax and accounting.
  • •For task-credit grants, including monthly membership allowances, we keep the record id, your numeric GitHub account id, invoice id when there is one, amount, kind of credit (a monthly allowance, a credit block or manual credit), when it stops being spendable, the project and month it belongs to, operator attribution and date. We also keep each amount reserved, spent, released or reversed against it, with the task it was for, and the granted and used balance totals. These records remain linked to your account after account deletion so we can explain the balance and prevent the same invoice from recording an adjustment twice. We remove the free-text operational note when you delete your account. The retained accounting fields have no automatic expiry period.
  • •Paid membership records retain your numeric account and project identifiers, Stripe customer, Checkout Session, subscription, invoice, payment and refund identifiers, amounts, billing periods, status and event history. We retain dispute freezes and accounting alerts, including credit owed after a refund. These records remain linked to your account for accounting and to prevent duplicate billing or credit, with no automatic expiry. Subscription Checkout sends Stripe opaque billing identifiers and the plan price, not your repository, project brief or code. Cancel and finish any open membership billing before deleting your account; an open subscription or Checkout prevents account deletion so recurring charges do not continue after access is removed.
  • •Your answers on the audit quote page are deleted, and so is their copy on any audit order you placed; the order keeps its price and line items, because we keep payment records for tax and accounting.
  • •Your project record and delivery policy, their revisions and confirmed snapshots, repository list, services map, team proposals and their resolutions, notes, questions and operator-recorded membership history are deleted with your account. Deleting a project record also deletes those associated records. Copies already committed to your repository or attached to a task are separate copies; this does not remove files from GitHub or a developer's local checkout.
  • •Every MCP connector you created, so a token minted for one stops working.
  • •Every team connector you created, deleted with your account the same way as your MCP connectors, so a token minted for one stops working. Proposals, notes and questions you wrote on another owner's project stay in that owner's record with the author shown as “former operator”. A draft you prepared for an owner shows the same label.
  • •A sign-in or CLI token from before you asked us to delete your account can't claim paid tasks afterwards. While deletion is running it gets “Your account is being deleted, so you can't claim new paid tasks.”, and after that “This account has been deleted, so you can't claim new paid tasks.”

8. Your rights

EU/EEA & UK (GDPR / UK GDPR)

Subject to applicable law, you have the rights of access, rectification, erasure (“right to be forgotten”, Art. 17 — including via terminalhire sync --delete, which hard-deletes your synced profile), restriction of processing, data portability, objection to processing, and the right to withdraw consent at any time without affecting processing already carried out. You also have the right to lodge a complaint with your supervisory authority.

California (CCPA / CPRA)

If you are a California resident, you have the rights to know, access, correct, and delete your personal information, the right to opt out of the sale or sharing of personal information, and the right not to be discriminated against for exercising these rights.

We do not sell your personal information — with one possible exception you control: the consented, developer-initiated lead share that you explicitly approve to a named buyer. [to be finalized with counsel: counsel to confirm whether the consented lead share constitutes a “sale” or “sharing” under CCPA/CPRA and update this disclosure and any opt-out mechanism accordingly]. We never sell or share the anonymous or on-device data described above.

9. Automated decision-making & profiling

TerminalHire does not make solely-automated decisions that produce legal or similarly significant effects about you (GDPR Art. 22). Job matching is a suggestion tool: it matches public job postings to your skills using a profile that never leaves your device, and you decide whether to apply. This profiling has no legal effect and never determines an outcome on its own.

10. Security

  • •Your local profile is encrypted at rest with AES-256-GCM.
  • •Data in transit is protected with TLS.
  • •GitHub access is minimized to the read:user scope (public data only).
  • •Employer-repository sessions are excluded from profile building by default.
  • •Server-side Tier-1 data is stored in Neon with encryption at rest.

No system is perfectly secure, but we work to protect your data using measures appropriate to the risk.

11. Children

TerminalHire is not directed to, and is not intended for use by, individuals under the age of 16. We do not knowingly collect personal information from children. If you believe a child has provided us data, contact us at privacy@staqs.io.

12. How to exercise your rights

Email us at privacy@staqs.io. For the data you control directly, you can also use the CLI:

  • •terminalhire sync --delete — hard-delete your server-side Tier-1 profile (erasure).
  • •Delete your local installation and ~/.terminalhire/ directory to remove all on-device data.

13. Changes to this policy

We may update this policy from time to time. When we do, we will revise the “last updated” date and, where required, provide a more prominent notice. Material changes will be communicated as required by law.

Effective date: June 15, 2026 · Governing law / jurisdiction: the laws of the State of Delaware, USA, without regard to its conflict-of-laws principles · This draft is pending legal review and is not legal advice.

Project feedback and sources can include photos, video, audio, PDF and text files. Shared attachments are visible to current project members. A developer who holds a live claim on a task can also read, for that task only, the owner’s change requests on that claim and the notes the owner added to the task or to that claim, with their attachments. On a task that requires the owner’s approval of each developer, the notes and their attachments are readable only after the owner approves that claim. The developer reads them, and the task’s links, on the claim page; the credential their CLI saved can read the same things, and the same rule decides both. That ends when the claim is released or rejected. It does not grant access to another task’s notes, to a collaborator’s notes, to project-level notes or to any other project file. Recording starts only after you choose to record. Attachments expire after 180 days and unsent uploads after seven days. Deleting your account removes your authored files and queues their removal from storage.

A task draft your assistant writes through the connector can hold notes it took from your conversation with it, alongside the task text and links it sends. You can add files to that draft on its confirmation page. When the draft is posted, the notes and files become a note on the task, which project members and the developer who claims the task can read. A file added to a draft that is never posted is an unsent upload and is removed after seven days.