PRIVATE BETA 4 of 8 seats open · Request access →

The agent trust network starts with visibility.

TapPass is the runtime governance layer that turns every AI agent call into a classified, scoped, auditable action. So risk, legal and the board can finally trust what's in production.

Start for free Join the private beta

14h agoPII

sales-outreach tried to enrich a customer record with an external API

claude-opus-4 · 3rd attempt today

SlackOps

"Who owns hiring-screener? It's shortlisting on gender-correlated features."

#ai-risk · posted 2m ago

Audit requestDPO

"Can you produce the full log of every agent that saw client X's data in Q1?"

Regulator · reply by Friday

AlertPrompt injection

support-triage executed a tool call it shouldn't have access to

GPT-4o · 08:41

FinanceApproval

Refund of €8,200 auto-issued without human sign-off

finance-approver · yesterday

BoardQ1 review

"Who actually owns AI risk this quarter?"

CRO → CISO → ?

eu.tappass.ai / agents

Agents

12 agents · 4,812 requests · 94 blocked

Live

AgentTierTraffic 24hStatus

support-triageclaude-sonnet · customer ops

Tier 1Healthy

hiring-screenergpt-4o · people

Tier 32 held

sales-outreachclaude-opus · sales

Tier 2Blocked

contract-reviewermistral-large · legal

Tier 2Healthy

The business risk of ungoverned AI isn't theoretical.

1 in 3

AI pilots stall before production, blocked by legal, risk or brand teams who can't sign off on what they can't see

8 weeks

Typical delay before a working copilot reaches customers once a single sensitive-data incident is raised

€40M+

Deal size lost when enterprise buyers ask for an AI governance artifact you can't produce on time

Your team is chasing AI incidents your platform should already be handling. That ends now.

Every Slack thread, every panicked audit request, every manual log-pull. All of it is work TapPass was built to run. Inline, every call, no heroics.

DPO9d ago

Re: Missing audit log

Still waiting on the full log of every agent that touched a specific case. Could someone please pull this before Friday?

Head of Legal8d agoURGENT

Hiring AI: bias concern

Legal flagged that the hiring-screener auto-ranks on a feature we can't justify under the AI Act. What's our rollback plan?

Head of AI RiskSlack · 14:36

#ai-risk

Who owns agent sales-outreach? Customer data just ended up in an external enrichment call. It has been sitting like that for 3 weeks.

Compliance lead6d ago

Fwd: Regulator reply

Sorry to bother again. The AP needs breach-notification evidence by Friday or they will escalate under Art. 33.

Platform lead2d ago

Re: Onboarding agent

Could you please verify this agent has the right capability scope before it goes into production?

2 days overdue

Reminder

Call back Audit prep · regulator response

snooze·dismiss

Board chair1d agoBoard

Quarterly AI review

Can someone tell me, in one slide, every AI agent running in production and who is accountable for each one?

CISOjust now

Re: Go-live decision for copilot v2

Can we actually trust this? Can it go live? I need a one-line answer and the evidence behind it before I sign the board memo.

Principles

Three commitments that make AI governance actually work.

Governance, not dashboards.

Every agent call is classified, scoped and enforced in-flight, not reviewed in a weekly report a week too late.

Inlineevery requestUnder 30msmedian added

Built for risk & compliance.

Policy is authored by the people regulators actually call. No engineering tickets. No waiting for a platform roadmap.

No-codepolicy editorFour rolesbuilt in

Evidence, on demand.

Every call leaves a signed, replayable trail. Breach notification, audit pack, regulator response: generated in minutes.

72-hourGDPR clockHash-chainedtamper evident

The layer in between

Your core systems store logs.
TapPass runs the governance.

Your SIEM, data warehouse and model providers are built to store.

Not to classify every prompt in-flight, enforce policy inline, scope identities, or stand up a regulator-ready breach response in 72 hours.

→Risk & compliance teams define policy, no IT tickets
→AI classifies every request and collects the missing evidence
→Every agent call is logged, scoped and tamper-evident
→Connects to your stack via one base URL and your SSO

Your systems stay the source of truth. TapPass runs governance on top, so your team no longer has to.

Agents & copilots

OpenAI · Anthropic · Mistral · Bedrock

12 in prod

TapPass governance layer

Monitor · Detect · Enforce · Harden · Respond

< 30ms

Audit record sealed

ClassifyScopeEnforceSeal

Your systems of record

SIEM · Data warehouse · CRM · SSO

Synced

Features

Everything you need to govern AI agents in a regulated enterprise.

Built around one governance layer. Configurable by risk & compliance. Powerful enough for the most sensitive agent workflows.

INC-2847Blocked

Agentsales-outreach

ClassifierPII · 0.88

OwnerCompliance lead

Evidencepayload hash · classifier trace

INC-2848Resolved

Agenthiring-screener

ClassifierBias · 0.71

OwnerDPO

Messy agent traffic → clean governed cases

Every request and response becomes a structured, owned case. Classifier scores, payload hashes, and decision trails are attached automatically. The case is ready to move forward, or be defended to an auditor.

Customer PII

0.88

Credential

0.72

Cross-client

0.64

Jailbreak

0.07

Data is automatically detected in every call

AI reads every prompt, tool call and response. PII, credentials, cross-client data and jailbreaks. Scored, logged and routed in real time.

No PII leaves the perimeterscope: all

Block

Hiring needs a humanEU AI Act Art. 14

Hold

Redact IBAN & IDsbefore log

Redact

Enforce what's missing

TapPass automatically requests human approvals, redacts sensitive fields, and blocks risky calls. Plain-language rules, shipped like code.

Read CVs

Call GPT-4o

Send emails

External APIs

See what each agent can (and can't) do

Scope identity, not credentials. Know exactly which tools, data and destinations each agent is authorised to reach, at a glance.

OpenINC-2847 · GDPR 3314h left

✓Blocked at gateway13:42

✓Notification drafted13:42

3DPO review & send···

When the clock starts, we run it

Every incident drafts its own regulator-ready filing: logs, classifier scores, evidence packaged. A human reviews and sends.

Head of AI Risk

CISO · 12 agents

Compliance lead

VP Plat · 4 agents

Review

Head of Legal

AI Risk · 7 agents

Route the work, to the owner

Every agent has an owner. Every incident has an escalation path. Every signal lands in Slack or Outlook, not another dashboard.

sandboxclaude code on jane's mac

WRITE/etc/hostsblocked

EXECrm rf ~/Documentsblocked

WRITE./out/report.jsonallowed

NETapi.stripe.comapprove

Local sandbox for coding agents

Every shell, file and network call your coding agent makes on the workstation runs through a per repo policy. Block destructive commands, require approval for outbound calls, keep credentials out of reach.

MCP server · tools exposed

read_customergovernedscope: tenant only, PII redacted on return

search_kbgovernedrate limited, query logged

refund_orderholdrequires human approval above €500

exec_shellblockdenied by policy, not in declared scope

MCP governance for every exposed tool

Agents expose capabilities to LLMs through MCP servers. TapPass sits at the MCP boundary, inspects every tool call, applies scope, redacts returns and writes the decision into the same audit trail as the rest of your agents.

Use cases

Built for the industries AI actually lives in.

Start with one team and one agent. Launch in weeks, not quarters. Expand as you grow.

Financial services

Inline governance for copilots in banking and insurance.

From first customer interaction to approval. Audit every tool call, protect PII before it leaves the perimeter, and stay DORA-ready.

refund approvalsKYC copilotsclaims triage

claims-copilot

TierTier 2

Today1,208 calls

Blocked11 · PII

DORAAligned

Healthcare

Keep patient data where it belongs. Govern every clinical assistant.

Classify PHI in-flight, enforce scope of use, and produce a defensible audit trail for every clinical agent.

clinical triageadmin copilotsPHI redaction

triage-assist

TierTier 3

PHIRedacted · 98%

ConsentPer-call

AuditSealed · EU

Public sector

AI Act readiness without an 18-month programme.

Map agents to the Act's high-risk categories, apply the right controls automatically, and export the compliance pack your auditor expects.

citizen intakegrant scoringhigh-risk mapping

intake-bot

Risk classHigh

Art. 14Human reviewer

Art. 12Auto-log

EvidenceExportable

Legal & professional services

Prove every AI action. Preserve privilege.

Keep privileged content out of third-party models, scope agent access by matter, and maintain a signed, tamper-evident log.

contract diffmatter scopingprivilege guard

contract-diff

MatterScoped

PrivilegeGuarded

Calls312 today

StatusHealthy

Integrations

One layer between your agents and everything they touch.

Plug into the model providers, frameworks and systems your teams already use. No rip-and-replace.

Model providers

OpenAI

Anthropic

Gemini

Mistral

Bedrock

Vertex AI

Groq

Cohere

Agent frameworks & coding agents

LangChain

LlamaIndex

CrewAI

AutoGen

Pydantic AI

Claude Code

Cursor

Haystack

SDKs for Python, TypeScript and Go · See integration docs →

Private beta · invite-only

TapPass is in private beta. Get the first build, not a waitlist email.

We're running a small, hands-on beta with risk, legal and platform teams in banking, insurance and healthcare. No pretend customer quotes. If you're piloting AI agents in production, we'd rather get you on the product.

1
Weekly build, direct to younew features and policy packs every Friday, shaped by this cohort
2
Shared Slack with engineeringyour DPO, risk and platform lead in the same room as the team building it
3
Two-week integration, your EU regionno roadmap tickets, no rip-and-replace. We deploy where you deploy
4
Locked pricing through GAbeta pricing grandfathered, named support, priority on the public launch

Beta cohort 4/8 seats filled

Designed for European risk teams

Built in Europe, shaped by the frameworks your DPO already reads.

Every control below is built against the articles your risk, legal and audit teams live with, so the evidence your agents produce lines up with the documents your regulator expects.

GDPR

Breach response, drafted for humans

Incidents build a notification draft, classifier trace and evidence bundle, shaped against Article 33 and refined with the beta cohort. Your DPO still signs off.

EU AI Act

Controls mapped to the articles

Policies line up with Articles 9, 12, 14 and 17 so the artefacts you configure speak the language your auditor expects. Full high-risk pack lands with GA.

NIS2 · DORA

Sealed audit, scoped identity

Hash-chained logs and per agent capability scoping are live today. ICT third-party reporting lands with GA, shaped by this cohort.

EU deployment

Private, set up with you

Designed to run inside your EU region. Beta cohort deployments are hands-on with our team; self-serve lands with GA.

From first agent in production to governance that runs itself.

Start with one agent. Launch in weeks, not quarters. See the difference. Expand as you grow.

Start for free Talk to the team