Security & privacy
Pushary is built on data minimization. To ask you to approve or answer, it needs the question and the tool name, nothing more. Your code and files stay on your machine, and your transcript does too unless you turn on transcript sync, which is end-to-end encrypted. Here is exactly what that means.
A lightweight hook runs in your agent (Claude Code, Codex, Cursor, Hermes, or any MCP client) when a tool call needs a decision. The agent and your code stay on your machine.
Pushary forwards the question and the tool name to its API, applies your per-tool policy, and if a human decision is needed, sends a push to your phone. Your code is never part of that payload.
You approve, deny, or answer from your lock screen. The decision page links are HMAC-signed so they cannot be forged.
Your answer returns to the hook and the agent keeps working. The question, tool, and your decision are written to an audit trail, never your code.
Pushary has two data paths, and they are encrypted differently on purpose. The decision path is always on and deliberately minimized rather than end-to-end encrypted: it carries only the question and the tool name, so there is a server-side, queryable, exportable audit trail of who approved what, which teams and compliance owners need and an end-to-end-encrypted tool cannot keep. Transcript sync is optional and off by default. When you turn it on, your transcript is end-to-end encrypted on your device with a key only your device holds, and Pushary stores ciphertext it cannot decrypt or read. Traffic is over HTTPS, decision links are HMAC-signed, and pending questions expire from cache on a short timer.
No. Pushary sends only the decision: the question text, the tool name, and minimal notification metadata. Your source code, file contents, diffs, and secrets never leave your machine, and your transcript stays local too unless you turn on transcript sync, which is end-to-end encrypted. The agent runs locally; the hook only forwards the prompt that needs a decision.
Partly, and by design. Pushary has two data paths. The decision path is always on and deliberately minimized rather than encrypted: it sends only the question text and the tool name, so there is a queryable, exportable audit trail of who approved what, which an end-to-end-encrypted tool cannot keep. Transcript sync is optional and off by default. When you turn it on, your transcript is end-to-end encrypted on your device with a key only your device holds, and Pushary stores ciphertext it cannot read. Traffic is over HTTPS.
Pending questions are held briefly in Redis with a short time-to-live and then expire. The audit trail records each question, the tool involved, and the human decision (who decided, when, and under which policy), so you and your team have a record. It does not store your code.
No. Decisions are scoped to your account and site, and the decision page URLs are signed with an HMAC secret, so a link cannot be forged or replayed by someone else.
Approve from your phone, set guardrails once, and keep an exportable audit trail. Works with Claude Code, Codex, Cursor, Hermes, and any MCP client. 7-day free trial.
