Verify in your environment.
A two-week, zero-cost deployment into your AWS or GCP account. Test deterministic enforcement, SIEM integration, and cryptographic auditing on real developer traffic. If it fails your acceptance criteria, terraform destroy removes it cleanly.
What the pilot costs you: roughly a few tens of dollars a month in your own cloud infra, plus 25 free seats. What it's worth: being able to answer the auditor's "what left through AI tools?" question — which today takes weeks of interviews and still has no real answer.
Full production capabilities.
Single-tenant control plane
Deployed exclusively in your GCP or AWS environment. We have zero access to your prompt data or audit logs.
Your own hostname, your TLS
A custom domain with managed certificates, terminating on your infrastructure — not ours.
Sign in with your IdP
SAML or OIDC single sign-on — your team uses the credentials they already have.
Seats from your directory
Operators created automatically from your directory, each one's role following group membership. Every add and removal is audited.
Operator and admin access
Two operator seats and one admin to start, more on request. Credentials live in your own secret store — never on a Tether server.
One-line developer install
A single command per machine. Activity starts flowing into the console — who, which agent, which host — within minutes.
Streams to your SIEM, day one
Splunk, Sentinel, or syslog — all live in pilot. Plus posture export for Okta and Entra, and an append-only audit log you can pull on a schedule.
Signed webhooks for the rest
Verifiable outbound events for any downstream system — SOAR, ticketing, your own pipeline.
Provisioning detail, routes, and signing schemes are in the technical brief →
What support you get
- A shared Slack channel with a Tether engineer
- A Tether engineer on your first policy publish
- Optional daily check-in through the two weeks
Five things you can
check by day 14.
If we cannot hit these by day 14, the pilot has not delivered. Mark the box; do not move the goalposts.
- At least 3 developer machines connect from the one-line installer, each reporting posture at startup.
- At least one approval request reaches an operator and is resolved inside the target window (default five minutes).
- At least one advisory finding is promoted into the deterministic floor, so it blocks on future requests.
- A policy publish and revert round-trip is visible in the audit log.
- An allowed-to-blocked transition is auditable when a developer's prompt contains a customer-specific pattern you've added.
What you'll see and feel in week one.
Five moments — the same loop production runs every day — on your own traffic, in your own console. You can watch the full demo first.
A developer uses an AI tool — and you can see it
The floor catches what you named
A risky moment pauses for a human
An operator decides in seconds
The developer unblocks — and you have evidence
14 days. Concrete activity per day.
| Day | Activity |
|---|---|
| Day 1 | Tether provisions the tenant. Customer adds DNS records. Customer uploads IdP metadata. |
| Day 2 | Customer admin mints invite codes. Two operators sign in via SSO. First developer connects via BYO installer. |
| Day 3-4 | Customer authors three regexes for sensitive patterns specific to the org. Publishes policy. Verifies audit entry. |
| Day 5-7 | Real developer traffic flows. Operators triage judge reviews. At least one promote-to-blocklist. |
| Day 8-10 | JIT loop tested: a developer asks to paste credentials, operator denies, audit entry recorded. |
| Day 11-12 | Customer reviews the policy audit export. Verifies posture reports from every connected machine. |
| Day 13-14 | Debrief against success criteria. Decision: extend, expand, or unwind. |
We'll tell you the limits
before we provision anything.
A pilot build has documented edges — single-tenant isolation, shared install tokens, advisory workspace enforcement, best-effort SLA, and a script-based installer rather than an MDM package. None of them is a surprise we spring later. If any is a hard gate for your environment, we'll say so on the first call, and the full list is in the technical brief →
"Don't we already have this?"
Most security teams already run at least three of EDR, CASB, IdP/SSO, and DLP. Tether replaces none of them. It covers the one gap each of them misses: what your developers send to AI tools. Here's how they line up.
Full 13-row stack-fit map (Tether vs EDR / CASB / IdP / DLP)
| Capability | Tether | EDR | CASB | IdP / SSO | DLP |
|---|---|---|---|---|---|
| Sees prompts / completions before they leave the IDE | ✓ | — | — | — | — |
| Attributes activity to a specific coding agent | ✓ | partial | — | — | — |
| Inline JIT prompt approval (clipboard / download / external link) | ✓ | — | — | — | — |
| Policy distributed straight to the developer machine (no MITM) | ✓ | ✓ | depends | — | depends |
| Operator console scoped to developer / AI-tool activity | ✓ | — | — | — | — |
| On-device payload inspection before TLS† | partial | ✓ | — | — | partial |
| Air-gap / no-cloud-dep pilot | ✓ | depends | — | depends | depends |
| Cross-IDE policy (VS Code, Cursor, JetBrains, CLI agents) | ✓ | — | — | — | — |
| Process-level egress block | partial (proxy-based) | ✓ | — | — | — |
| Endpoint malware / behavioral threat detection | — | ✓ | — | — | — |
| SaaS shadow-IT discovery and posture | — | — | ✓ | — | — |
| User identity and SSO federation | partial (via Jackson) | — | — | ✓ | — |
| File-level content classification across the org | — | — | — | — | ✓ |
† On-device payload inspection before TLS: Tether reads request bodies on the configured-upstream HTTP path. TLS-tunneled traffic is host-classified, not body-inspected — see the technical brief · scope & threat model.
The short version: Tether handles the AI tools on the developer's machine. The other four still do the jobs you bought them for. If your EDR, CASB, IdP, or DLP vendor ships a credible AI-tool story later, that's real competition — ask us about it directly rather than take an analyst's word for it.
One terraform destroy
and the tenant is gone.
Tearing down the pilot is one terraform destroy. Your data lives in your own Secret Manager, GCS, or EFS — deleting the tenant deletes its state. There's no Tether-hosted copy to delete, because the whole deploy is in your account.
$ cd deploy/gcp/terraform
$ terraform destroy -auto-approve
# Cloud Run service, GCS bucket, Secret Manager secrets — gone.
# Uninstall the proxy and daemon on each developer machine.
$ tether-proxy uninstall && tether-workspace uninstall
Two weeks, one decision.
Tell us your cloud (GCP or AWS), your IdP, your SIEM, and the AI tools in scope. We bring a tenant ready to provision.