Enterprise / Trust / Security

SecurityEarly access

Security you can check, not just read about.

Isolation enforced by the database, credentials encrypted at rest, a signed audit chain, and every guard tested by breaking it on purpose.

Fig.1 · sample
runtoolruleverdict
run_8f25pull_request.openfactory_scopeconfirmed
run_8f24invoices.listread_onlyconfirmed
run_8f23whatsapp.sendservice_windowunknown
run_8f22bookings.cancelambiguous_targetrefused
run_8f21payment_link.createsend_to_customer → askconfirmed
run_8f20bookings.createbooking_hoursconfirmed
■ confirmed 4■ refused 1■ blocked 0■ unknown 1done-without-check 0

How it works.

  1. 01 /

    Isolate

    Each workspace fenced by row-level security, with least-privilege roles.

  2. 02 /

    Govern

    One gate, decision tables and exact approvals for every action.

  3. 03 /

    Prove

    A signed, tamper-evident record of what ran and why.

What you get. Each point labelled with its real status.

1.1Early access

Sabotage-tested guards

A check counts only after it has been seen to fail when broken.

1.2Early access

Encrypted credentials

Encrypted at rest, scoped per workspace.

1.3Coming

Hardening in progress

Signed app proxies, origin allowlists, SSRF-safe fetching and route audits in CI.

1.4Coming

MCP security gateway

Authenticated, identity-bound calls with pinned tool descriptions.