Operational AI governance

Your rules.
AI follows.

Turn policy and regulatory requirements into controls enforced across models, agents, and tools, with evidence for every governed decision.

Start free Take free assessment
Works with your existing proxy and providers EU-hosted Fail-closed by default
Governed request, in path req_8f21c40b · 17:05:29
Agent identified trading-analyst
Organisation policy inherited High-risk agent baseline
Sensitive data detected 2 entities redacted
Model selected obstruo-chat-eu, approved
Tool requested place_order
Decision Blocked and recorded
Policy version, decision, routing, redaction and tool call written as one evidence record
Agent record GOVERNED
OWNERTrading platform team
RISKHigh
DATA CLASSIFICATIONConfidential, personal data
APPROVED MODELobstruo-chat-eu
PERMITTED TOOLS4 of 11, read only
ACTIVE POLICIES6 inherited, 1 project
EVIDENCEComplete
Neutral by design. We do not build models, we govern them. obstruo sits in front of
OpenAI Anthropic Google Mistral Azure Bedrock
Define, enforce, prove

One policy. Active across every AI project.

Your organisation sets the rules once. obstruo applies them to every project, agent and request, and keeps the evidence that it did.

01 Define
Set organisation policy

Locked, recommended and optional policies, inherited by every project the moment it is created.

POLICY CATALOG
EU personal data redactionLOCKED
Approved models onlyLOCKED
High-risk agent baselineRECOMMENDED
Per-project spend ceilingOPTIONAL
02 Enforce
Apply it in runtime

Redaction, model restrictions, guardrails, budgets and tool permissions act on the request itself, not on a report.

REDACTION STANDARD, EU
Client {PERSON_18b}, IBAN {IBAN_3d2}
Guardrails3 active
Tool permissions4 of 11 allowed
Unevaluated policyFail closed, selectable
03 Prove
Preserve the evidence

The policy version, the decision, the routing, the redactions and the tool call stay connected as one record.

AUDIT GRAPH, req_8f21c40b
policy v14inherited
redaction2 entities
routingeu-central
tool callblocked
Linked to project, vendor and incident
Scope of control

Control what AI can see, use and do

Protect data

Redact PII and secrets before they reach models, and again when they return through tools.

Govern models

Control which providers and models are allowed, behind stable logical model names.

Govern agents and tools

Apply policy by agent identity, and restrict which MCP tools and actions each one can reach.

Control operational risk

Enforce budgets, rate limits, safety guardrails and fail-closed behaviour when a check cannot run.

Maintain evidence

Connect runtime decisions to projects, agents, vendors, incidents and approvals.

One policy surface for all five

Data, models, agents, risk and evidence are governed from the same place, so a rule changes once and applies everywhere.

See it from both sides
One record, two audiences

Compliance and engineering see the same reality

No reconciliation meeting. A single blocked request appears in both views, described in the language each team works in.

req_8f21c40b trading-analyst, 17:05:29
app.obstruo.ai/monitoring/req_8f21c40b 403 BLOCKED
REQUEST
POST /v1/chat/completions
key: trading-prod, stream
MODEL
obstruo-chat-eu
resolved: provider eu-central
REDACTION
2 entities
PERSON, IBAN
GUARDRAIL
3 passed, 0 failed
injection, jailbreak, egress
TOOL CALL
place_order, denied
not in permitted tools
LATENCY
214 ms
19 ms added latency
app.obstruo.ai/compliance/evidence/req_8f21c40b EVIDENCE COMPLETE
APPLICABLE POLICY
High-risk agent baseline
v14, organisation-wide, locked
AGENT AND ASSET
trading-analyst
Asset A-114, risk high
OWNER
Trading platform team
Accountable: Head of Markets IT
VENDOR
EU model provider
DPA on file, EU processing
INCIDENT
INC-2041, open
Unpermitted trade action
APPROVAL
Model approved
Signed off by compliance, March 2026
FRAMEWORK MAPPING
EU AI Act, Article 14GDPR, Article 32ISO 27001, A.8.16SOC 2, CC6.1
Where teams start

Three use cases that need governance first

See all nine use cases
Govern production AI agents

Once an agent can place an order or issue a refund, tool permissions become the control that matters. Allow the reads, deny the writes, record the attempts.

Agent identity Tool permissions
Protect sensitive data

Personal data and secrets reach models through prompts, retrieval and tool responses. Redaction runs in path, in both directions, and records what it removed.

Redaction Secret detection
Control models and providers

Applications ask for a logical model name. Policy decides which approved provider and region answers, and refuses the ones your review has not cleared.

Approved models Residency
Compatibility

Fits the AI infrastructure you already operate

You do not have to replace your proxy or re-onboard every project to start governing AI.

option 01
Use obstruo directly

Point your applications at obstruo and let it hold the provider connections, the policy and the evidence.

option 02
Chain it with your proxy

Keep the proxy you already run for traffic, and put obstruo in front of it for policy and evidence.

option 03
Apply it selectively

Start with the higher-risk projects and agents, then widen coverage as the policy proves itself.

Capability status

What is live today, and what is not

Everything below is labelled honestly, so your security review does not have to guess.

LIVE
Redaction Guardrails Models and routing MCP governance Policies Audit Incidents Approvals
PRIVATE PREVIEW
Agent registry

Available to design partners. Ask us for access.

IN DEVELOPMENT
Human oversight Exposure map Kill switch
PLANNED
Browser adapters Coding-assistant adapters Microsoft Copilot adapters Custom PII and guardrail models Skills Credential vault More evidence packs
Security and trust

Built to the standards your auditors ask about

EU AI Act
Article 14 oversight records
GDPR
Article 32, minimisation in path
SOC 2 Type IIPLANNED
Controls in place, audit not yet completed
ISO 27001PLANNED
Aligned to the controls, certification in progress
EU-hosted
EU regions, pinned per project
Failure behaviour
Fail closed or open, your choice
Questions we get in security review

Frequently asked

Search all 47 answers
Does obstruo replace our AI proxy?

No. Use it directly or chain it in front of the proxy you already run. Either way, policy and evidence stay in one place.

Where is data processed?

EU-hosted by default, in EU regions only. Traffic can be pinned to in-jurisdiction endpoints per project.

What happens if a control cannot be evaluated?

You choose the behaviour per policy: fail closed and refuse the request, or fail open and let it through. Either way the outcome is recorded with the policy version that caused it.

How much latency does enforcement add?

Around 20ms p50 for the enforcement path. Redaction with entity detection adds more, and is measured per project.

Do we have to onboard every project at once?

No. Most teams start with two or three higher-risk projects, then inherit the same organisation policy as coverage widens.

What evidence do auditors actually get?

Policy version, decision, routing, redaction counts, tool calls, incidents and approvals, exportable per project, agent or framework.

Your AI. Your data. Your control.

Start free on one project, then make the same policy executable across the organisation.

Start free Take free assessment
Free
€0
One project, evaluate the controls
Teams
from €149/mo
Production projects and agents
Enterprise
Custom
Organisation-wide, sovereign deployments
See full pricing