OPEN SOURCE · APACHE-2.0

Open-source AI-native privacy ops for DSAR workflows

Governed AI agents that help privacy teams intake, search, reason, evidence, and approve DSAR work — without losing auditability or control.

View the Demo →★ Explore on GitHub
✓ Human-in-the-loop by default
✓ Full evidence & audit trail
✓ No side effects without approval
instrata · operator
LIVE
DSAR-2041
Right to Know
SLA · 6d left
Dana Whitfield · dana.w@example.com
California (CCPA) · identity verified
Agent activity
›
search() across 4 source systems — 3 matches
›
Matched on email + last name in Billing & CRM. Marketing match is low-confidence.
Billing · 1CRM · 1Support · 1Marketing · low
Proposed action · Compile disclosure package
Awaiting operator approval — no export until approved
GATEApprove
Built for:CCPA / CPRAGDPRRight to KnowDelete My DataCorrectionAccess requests
The problem

DSAR work is slow because the data is everywhere

Every request means collecting intake details, verifying identity, searching fragmented systems, deciding which records actually match, tracking regulatory deadlines, and preserving evidence for audit — mostly by hand.

45 days
Typical statutory clock, with hard penalties for missing it.
6–12
Disconnected systems a single record can live across.
100%
Of actions must be defensible and evidenced on audit.
How it works

From consumer request to approved action

The agent does the searching and reasoning. A human always makes the call before anything leaves the building.

1
Consumer
Request is submitted
A consumer files a privacy request through the intake portal — request type, identity, jurisdiction.
2
Operator
Lands in the dashboard
It enters the operator queue with status, SLA clock, type, and a verification check.
3
Agent
Agent searches sources
Governed tool calls query connected systems. Every call is visible and logged.
4
Agent
Explains match reasoning
The agent shows why each record matched — and flags the ones it isn't sure about.
5
System
Evidence is captured
Inputs, tool calls, results, and decisions are written to an immutable audit trail.
6
Human gate
A human approves
Emails, exports, and deletions happen only after an operator approves the proposed action.
Capabilities

Everything a DSAR needs, in one governed workspace

01
DSAR intake portal
Branded consumer-facing form that captures request type, identity, and jurisdiction.
02
Operator request dashboard
A queue with status, SLA clock, type, reviewer, and verification at a glance.
03
Source-system search
Governed queries across connected billing, CRM, support, and marketing systems.
04
Agent tool-call visibility
Every tool the agent calls is shown inline and written to the log.
05
Match reasoning
Plain-language explanation of why each record matched — and confidence.
06
Audit trail & evidence log
Immutable record of inputs, calls, results, and human decisions.
07
SLA & status tracking
Deadline clocks and status transitions on every request.
08
Approval-gated actions
Exports, emails, and deletions require explicit human sign-off.
09
Human-in-the-loop controls
Operators can edit, reject, or request changes to any proposed action.
10
Open-source architecture
Inspect the orchestration, audit the tools, add your own connectors.
Governance

The agent assists. It never acts on its own.

Every side effect — an email to a consumer, a data export, a deletion — is approval-gated. Operators inspect the reasoning, the evidence, and the exact proposed action before anything happens.

✓
Inspect before approve — full reasoning and evidence for each proposed action.
✓
Approval gates — block all external side effects until a human signs off.
✓
Immutable audit trail — records who decided what, when, and on what basis.
audit_trail · DSAR-2041
09:42intakeRequest received · Right to Know · CA
09:42verifyIdentity verified via email + DOB
09:43toolsearch() → 4 systems, 3 matches
09:43reasonMatched on email + last name (high)
09:44gateAwaiting approval · compile disclosure
Live demo

See a Right to Know request, end to end

Walk a real request from intake through operator review — agent search, match reasoning, evidence capture, and the approval gate before any action fires.

Launch the demo →Try the intake form
01Intake & identity
02Agent search & reasoning
03Operator approval gate
Open source

Transparent, inspectable, extensible

Read the orchestration logic, audit the agent tools, and add your own source-system connectors. Nothing about how decisions are made is hidden behind a black box.

★ Explore on GitHubRead the architecture
// register a source-system connector
registerConnector("crm", {
search: async (identity) => { … },
sideEffects: "gated", // never auto-run
audit: true
});
→ connector visible in operator tool-call log