Ripcord
THE GATEWAY
How it worksConnect your agents, pick a channel, the scorecard decides Control roomWatch a $48,200 wire get caught, live Launch an agentPick a brain, give it a job, set its rules. Leash included
THE RULES
Where the score comes fromThree actions walked end to end, every point on the table Risk policyThe JSON your rules compile to Your dataTokenized at the gateway: the model sees pseudonyms, never customers
WHERE THIS GOES Why it compounds every verdict prices the next Gate today, insure tomorrow the actuarial table no insurer has
BEFORE IT RUNS
Hold for approvalAbove the risk gate, payments fail closed. Silence means no Test-sized canaryAt high stakes a probe proves the rails before the real wire Owner routingApprovals land with the corridor owner, enforced by your org chart TokenizationPII becomes vault tokens before the model reads it Know your fleetEvery agent, tool, server and person the gate has seen. Dormant keys called out
AFTER IT RUNS
Snapshot and rollbackEvery destructive call snapshots first. Undo stays one click Batched sendsMass sends leave in batches. Stoppable at #50, not #2,000 RIPRest in peace: the kill switch. A tripwire catches the loop, everything after fails closed Conduct gradesEvery agent graded A to D on what it did, not what it claims What burns for nothingLoops, failed runs, idle agents: wasted turns counted, priced from your bill
THE MATH Expected-loss gating P × exposure × (1 − recoverability) CUSUM velocity windows structuring dies at payment two
BY CORRIDOR
Money$48,200 wired to a brand-new vendor. Held until a person owns the yes Data1,240 customer records deleted overnight. Snapshot first, rollback in one click Comms6,200 customers emailed by mistake. Halted at email 50, not 6,200 PublicThe launch post fires with last month's discount. Pulled in two minutes
WHO SIGNS
The blocked rolloutThe gate answers the question that stalled the agent The approval loadSafe actions run instantly. Only the risky ones reach a human AuditorsISO/IEC 42001 and NIST AI RMF, plus a hash-chained ledger
AND THEN Insure the fleet · roadmap every gated action builds the actuarial table The cost of waiting what the standoff burns per year, from your sliders
THE EVIDENCE
The incident fileTen sourced incidents, and the base rate behind them The Ripcord IndexConduct grades from real actions, and how they are computed The cost of waitingWhat the standoff burns per year, live from your sliders
FOR YOUR AUDITORS
Compliance mappingISO/IEC 42001 and NIST AI RMF, control by control Coverage mapWhat we can gate today, per agent, verified against vendor docs StatusLive checks from your browser, no synthetic history Near-Miss ReportWhat the gate caught, live from the demo workspace The weekly reportAgents ranked, graded, and the fixes spelled out, every Monday
STILL SKEPTICAL The questions skeptics ask fourteen answers, no hedging
Pricing Blog About
Book a demo Open Ripcord
Privacy policy · effective 8 September 2026

What we collect, and what we refuse to.

This policy covers the website at tryripcord.com and the Ripcord application at app.tryripcord.com (together, the Service), operated by Ripcord (the company being formed by Jeremy Hazan; the legal entity will be named here once incorporated). Questions go to [email protected].

The website

The website is static pages served by Vercel. We run no analytics scripts, no advertising pixels and no tracking cookies on it. Vercel and Cloudflare keep standard server logs (IP address, user agent, requested page) for their own operation; see their privacy policies for retention. The website sets no cookies of its own, so it shows no cookie banner: there is nothing to consent to. Two pages embed a third party for a function you ask for: the demo page loads a Cal.com booking calendar, and the pricing page's waitlist form sends the address and fleet size you type to our application. Cal.com may set its own cookies inside its frame, under Cal.com's privacy policy. If you email us, we keep the email.

The application: your account

When you sign in we store your email address, your display name, the time of sign-in and a session identifier. Sign-in is by a one-time link sent to your email or by Google. With Google sign-in we receive your email address, whether Google has verified it, your name and your Google account identifier; we do not receive your password and we ask for no other Google data. Your workspace is keyed to your email domain so that colleagues can join it.

The application sets one cookie: a session identifier, marked HttpOnly and SameSite, valid for thirty days and cleared when you sign out. It is strictly necessary to keep you signed in, so it needs no consent. The sign-in page loads Google's sign-in widget, which runs in its own frame under Google's privacy policy. The application also keeps a few small preferences in your browser's own storage, such as whether you skipped setup or have read a brief; those never leave your browser. The public demo signs visitors into a shared read-only workspace with the same session cookie and stores nothing about them.

The application: what your agents do

Ripcord's purpose is to record and gate the actions of AI agents you connect. For each action we store: the agent's name, the tool called, the action's parameters as they left your environment, the risk verdict and its reasons, the decision a person took and their name and role, and whether the action ran, failed or was reversed. This record is a hash-chained ledger and is kept for the life of your workspace, because its purpose is to be an audit trail.

Parameters are processed by the Ripcord gateway inside your own environment before they reach us. Where you configure data rules, personal data in tool arguments and results is replaced with pseudonyms there, and the table that maps pseudonyms back to real values stays in your environment. We never hold your systems' credentials: the gateway does, on your side.

What we do not do

  • We do not sell personal data, and we do not share it with advertisers.
  • We do not train models on your ledger, your agents' actions or your decisions.
  • We do not read the contents of the systems your agents act on; we see the action, not the database.

Processors

We use a small number of providers to run the Service: Railway for hosting the application and its database, Vercel and Cloudflare for the website, Resend for delivering sign-in links and notifications, Google for Google sign-in, Cal.com for demo bookings, Slack or Microsoft Teams when you connect them for approvals, and Anthropic for model inference when you choose a Claude brain or turn on the morning brief. Each receives only what its function needs. The current list is available on request and will be published on this page when the application opens to customers.

Retention and deletion

Account data is kept while your workspace exists. Sign-in links expire after fifteen minutes and are single use; sessions expire after thirty days. The ledger is kept for the life of the workspace. When a workspace owner asks us to delete the workspace, we delete accounts, keys, rules and the ledger within thirty days, except where a legal obligation requires keeping a record, in which case we keep only what the obligation requires and tell you.

Your rights

You can ask for a copy of the personal data we hold about you, ask us to correct it, ask us to delete it, or object to how we use it. If you are in the European Union or the United Kingdom you have these rights under the GDPR and may also complain to your supervisory authority. Write to [email protected]; we answer within thirty days.

Security

Sign-in tokens, session identifiers and gateway keys are stored as hashes. Transport is encrypted. The ledger is tamper-evident: every entry is chained to the previous one by a cryptographic hash, and any alteration is detectable.

Changes

When this policy changes we update the date at the top and, for material changes, tell signed-in users in the application. Earlier versions are available on request.