FAQ
Questions teams ask before they adopt
How SignTrail decides, where it runs, what it collects, and how a rollout starts.
No. SignTrail sits in front of the signers you already run and checks each request before it reaches them. Nothing is replaced and nothing is migrated.
No. SignTrail never stores keys, never signs, and never reads your wallet keys. Keeping keys secret stays with your HSM or KMS.
Nothing gets signed. SignTrail is fail-closed: in enforce mode, if it can’t reach a decision in time, the signature doesn’t go ahead. In observe-only mode the signature proceeds and the outcome is recorded.
One call, placed right before the service signs or calls a node API. The rest of your code stays the same. Node.js services use a small client library, and other runtimes such as Rust connect directly.
Observe first. SignTrail starts in shadow mode, where every request is checked and recorded but nothing is blocked. The allowlist is drafted from about a day of real traffic. Enforce mode is never switched on automatically. After shadow mode shows zero rejects, your operations team changes the setting.
None. SignTrail opens no ports and sends nothing out. It records which programs ran, which files were opened, where connections went, and whether signing files changed, never file contents, payloads, or environment variables.
Where the request came from. HSM and MPC policies check who signs and what is signed, while SignTrail checks whether the request came from the expected program, what it opened and connected to in the seconds before, and whether signing files still match what was deployed. A request from a tampered process can look valid to the signer, and SignTrail rejects it before it is signed.
No. Each approval is sealed to that request and signer and works only once. Reusing it, or attaching it to another request, is rejected.
Every decision becomes a signed record linked to the previous one by hash. The verifier reports chain links and signature authenticity separately, and the Evidence Pack shows each decision with its reason and the policy that applied.
No. It checks whether the signing host acted outside the allowlist, not whether the amount or the recipient is right. Keep your existing approval and policy checks for that.
No tool can promise that. SignTrail reduces unsafe signing risk by rejecting requests that come from an untrusted path before they are signed. It doesn’t cover what happens outside the signing host, or tampering inside memory or the kernel itself. It does not claim the evidence caused the transaction; its scope is the execution record of the process that submitted the request.
It runs on the Linux signing host, next to your signing service: Ubuntu 22.04 or 24.04 on x86_64, Linux 5.8 or later. A small host is enough (2 vCPU, 2 GB of memory), and the alpha covers one signing host.
Yes, with a small layout change. If the signing service runs in a container, SignTrail stays on the host and the container connects to it, and tampered files inside the image are still caught. For Windows, macOS, or an HSM appliance, a Linux host sits in front of the signing call, and SignTrail checks that host rather than the device behind it.
Very little. In alpha measurements the decision itself takes about 64 µs at the median, and the alpha supports up to 100 requests per second per signing host. These are alpha figures, not a production guarantee.
Any. SignTrail doesn’t interpret the transaction: it binds the approval to the request itself, so the same check works for any ledger. The items needed for integration are agreed for your setup during onboarding.
Yes. An adapter connects to HSMs through the PKCS#11 standard. Integration with real hardware is validated in a PoC, in your environment. Either way SignTrail only decides, and the HSM signs as it does today, only on ALLOW.
About a day of observation, a policy review, and another day to confirm zero rejects. Your team then reviews the shadow results and the allowlist before switching to enforce mode. The overall schedule is set together after a short environment check.
They are caught before they can block anything. During observation a reject is only recorded, and each one comes with a plain reason, such as an unlisted file or an unexpected destination, so your team either adds it to the allowlist or investigates. Every policy change is replayed against past traffic to confirm zero rejects before it goes live.
You refresh the baseline. When the executable or bundle changes, the list of expected files and the policy are updated, otherwise the new files would be rejected as changed. The update is checked and replayed against past traffic before it goes live.
Anything outside the allowed execution path is blocked with REJECT, and policy cannot soften that. When everything is on the allowlist the request passes, and the default is ALLOW. HOLD and REVIEW are values a deployment policy can choose so that a pass is held or reviewed instead of signed right away. Anything other than ALLOW never reaches the signer.
Teams that move digital assets or run privileged onchain operations: exchanges, custodians and financial institutions, DeFi protocols, and wallet or infrastructure providers. Most start with one critical signing workflow, such as cold-wallet transfers or treasury moves.
Didn’t find your answer?
Tell us about your signing workflow and we’ll answer for your setup.
Stop Hacks Right Before Signing
The safest signature is the one that never gets executed when the path is wrong.
