Redacted on your server
The SDK redacts inside your own process, from requests and responses alike, before a log is batched for upload.
SECURITY & PRIVACY
Every part of Restless is designed to keep our customers and their data safe from the start.
Sensitive headers and fields are redacted on your server, before a log is uploaded.
Recovery hints pass a secrets scan and two separate AI reviews before an agent sees one.
Opening an end user’s requests or profile in the dashboard is recorded in an audit trail.
PII
Restless records every call to your API. Sensitive values are stripped on your own server, so they never reach ours.
The SDK redacts inside your own process, from requests and responses alike, before a log is batched for upload.
Authorization, Cookie, Set-Cookie, Proxy-Authorization, X-API-Key and X-Auth-Token headers are redacted on every call.
Keys like password, token, secret, ssn, creditcard and cvv are redacted in query strings and JSON bodies, however deeply they’re nested.
Add your own headers, body keys and query params in settings or in code. They add to the defaults, which always apply.
A redacted value keeps its length, and a longer one its last four characters, so you can tell two keys apart without holding either.
The SDK doesn’t send your callers’ IP addresses to Restless.
Opening an end user’s requests or profile in the dashboard is recorded in an audit trail, with their email stored as a hash.
AGENTS
An agent calling your API sees what your API sends it. What Restless keeps of that exchange goes through the same redaction as every other call.
Restless doesn’t rewrite your responses. It adds debug headers to every call and a debug block to errors, and the rest is exactly what your API sent.
The copy Restless keeps is redacted whether the sensitive value went out in the request or came back in the response.
Errors are grouped from the redacted copy. A fingerprint built from an error message drops numbers, emails, URLs and quoted values first.
RECOVERY
A recovery hint is written once per kind of error, not built per request, so a caller’s data has no way into it. Every hint is scanned and reviewed before an agent sees it.
A hint is saved against an error fingerprint, and every matching error gets that same text, word for word.
Values from the request or response, headers, query params and the caller’s identity are never inserted into a hint.
A hint’s docs link uses the route template, like /users/{id}, never the real id. The page it opens shows only what’s in your OpenAPI spec.
Restless checks an error for prompt injection before a model reads it, so a malicious request can’t steer what the hint says.
The model drafts from the method, status, route template, error reason, your OpenAPI operation, scrubbed support answers and the code that throws the error. Never full bodies, raw URLs or who called.
A scan catches emails, phone numbers, SSNs, card numbers, API keys, tokens, connection strings, internal hosts and IPs, and record IDs. It decodes base64, hex and URL encoding first.
When an agent hits an error, the response tells it how to fix it.
See every call to your API, with the patterns behind your errors surfaced.
Filter your logs and replay any request to see exactly what failed.
GET STARTED
Restless is tightly integrated with your API, so we install a small SDK. Don't worry, we walk you through every step of the way.
Nothing is uploaded without your permission.
A terminal animation shows the Restless API setup from introduction through account login.