The platform behind your AI transformation

Every product in the Ikigai Intelligence platform is designed for regulated banking environments. They work independently and together, growing with your AI maturity from Foundation through to full Autonomy.

CheckpointComing Soon

One gateway. Every model. Complete control.

A centralised LLM gateway that gives every team in your bank safe access to AI. Bring your own keys, route to any provider, and enforce Zero Data Retention on every call. Every interaction is logged in a structured, queryable audit trail. When compliance needs to know what your AI did and why, the answer is already there.

For banks taking their first step into AI. Safe, governed access for every team.

Stage: Foundation

Connect any LLM provider with your own API keys (BYOK)

Route requests across models with automatic load balancing

Trace every request and response for investigation and compliance

Enforce Zero Data Retention on every model call

Track usage, costs, and adoption across teams

Control access with team and role-based permissions

Checkpoint Gateway

0 requests0 blocked

Provider Routing

OpenAI

35%

Anthropic

28%

Google

22%

Audit Trail

Live

StatusTeamModelZDRLatency

Overpass

Build AI-native banking apps in minutes, not months.

A CLI framework that scaffolds production-ready banking frontends. One command generates a complete Next.js application with design system, authentication, typed API clients, Docker, tests, and AI coding agents. Works in air-gapped environments with no cloud dependency.

For engineering teams ready to build. Ship internal tools and customer-facing apps fast.

Try Overpass↗

Stage: Foundation

Scaffold a full-stack banking app with a single command

Run in air-gapped environments with no cloud dependency

Theme automatically from your existing bank website

Integrate with core banking APIs out of the box

Include AI coding agents in every scaffolded project

Eject anytime. You own the generated code

Terminal

Active Accounts

12,847

+3.2%

Pending Reviews

-12%

Total AUM

€2.1B

+0.8%

Monthly Volume

M. SchmidtDE89...4217Active

K. YamamotoDE71...8834Review

A. PetrovDE44...1190Active

Overpass AgentOnline

Ask the agent...

Dashboard/Wire TransfersNew

Initiate Wire Transfer

From

DE89 3704 0044 0532 0130 00

GB29 NWBK 6016 1331 9268 19

Amount

€25,000.00

Ref

INV-2024-0847

Validation passed

Submit Transfer

Recent Transfers

SWIFT: DEUTDEFF€50,000Completed

SWIFT: COBADEFF€12,300Pending

SWIFT: BNPAFRPP€8,750Completed

Lightspeed

Ship battle-tested financial products. Faster, safer, smarter.

An AI-powered development companion that embeds deep core banking expertise into every stage of your SDLC. Three specialist AI agents review every pull request. Starter kits scaffold production-ready code. Months of implementation compress into days.

For teams implementing core banking. Expert knowledge embedded in your development workflow.

Stage: Acceleration

Review every PR with three specialist AI agents

Retrieve domain knowledge dynamically during reviews

Check code against OWASP and PCI-DSS security baselines

Assess business alignment for every technical change

Generate starter kits with full test infrastructure

Score deployment risk and flag change management issues

Lightspeed Review

Council Active

Specialist

Consultant

Business

feat: add tiered interest accrualOpen

#247 · 4 files changed · +218 −12

📄interest_accrual.py+142−4

📄test_interest_accrual.py+68−0

📄template_params.py+6−6

📄supervisors.py+2−2

Awaiting Council of Reviewers...

Domain Specialist

Core Banking Expert

Reviewing: feat: add tiered interest accrual

✓Correct use of ACCRUED_INCOMING address

✓Tier boundaries handle edge amounts

!Consider calendar-aware day count (ACT/365)

✓Hook execution order is correct

“Solid implementation. Day count convention should be reviewed with product.”

Security Consultant

Cybersecurity Review

Reviewing: feat: add tiered interest accrual

✓No hardcoded secrets or credentials

✓Decimal precision prevents rounding exploits

✓Input validation on all external parameters

!Add rate limiting on manual accrual trigger

“Passes OWASP and PCI-DSS baseline. One advisory recommendation.”

Voice of the Business

Business Alignment

Reviewing: feat: add tiered interest accrual

✓Tiered structure matches product specification

✓Customer-facing rate display is accurate

iMarketing: highlight competitive tier rates

✓Regulatory disclosure requirements met

“Aligned with business requirements. Ready for UAT.”

✓

Approved. Ready to merge

3/3 reviewers approved · 2 advisories · 0 blockers

✓Specialist

✓Consultant

✓Business

Merge Pull Request

Knowledge Bases

Your institutional knowledge, always evolving, always agent-ready.

A framework for building living knowledge bases from your domain expertise. Ingest documentation, APIs, codebases, and team conversations. Synthesise them into structured knowledge that every AI agent can retrieve and reason with. Not a static wiki. A system that evolves as your teams learn.

For organisations building AI capability. Turn what your experts know into knowledge every agent can use.

Stage: Acceleration

Ingest from documentation, APIs, codebases, and conversations

Generate agent definitions for any IDE automatically

Power every agent with contextual knowledge retrieval

Evolve knowledge continuously as your teams learn

Work across Cursor, Windsurf, VS Code, and CLI agents

Capture lessons from every interaction automatically

Knowledge Base Studio

Ingest

Synthesise

Retrieve

Ingesting Sources

📄

Core Banking API DocsDocumentation

💻

smart-contracts/Codebase

📋

Product RunbooksConfluence

💬

Slack #core-bankingConversations

🔧

Payments API SchemaOpenAPI

📝

Incident RetrospectivesDocumentation

0 docs processed

Knowledge Graph

0 topics synthesised

Hook Lifecycle & Execution14 articles

Balance Addresses & Posting23 articles

Smart Contract Patterns18 articles

Payment Processing Flows11 articles

Compliance & Audit Trails9 articles

Incident Patterns & Fixes7 articles

Agent Retrieval

Live

@cb-dev

How do I handle pre-posting hooks for overdraft checks?

@cb-reviewer

What are the security requirements for balance modifications?

@cb-risk

Show me past incidents related to interest calculation errors

Smart Contract Studio

From product idea to dev-ready specification in minutes.

An AI-powered workbench for configuring, simulating, and specifying financial products. Describe a product, simulate its behaviour with a real financial engine, generate grounded BDD test specs, and produce a complete project manifest ready for your engineering team to build.

For product and engineering teams. Business requirement to dev-ready specification in a single session.

Try Smart Contract Studio↗

Stage: Acceleration

Configure financial products with interactive live simulation

Simulate with a financial-grade engine (tiered interest, calendar-aware)

Generate BDD test specs grounded in real test frameworks

Produce complete project manifests with CI/CD pipelines

Include AI agent definitions and coding commands

Refine iteratively with edge cases and compliance scenarios

Smart Contract Studio

Vault

Configure

Simulate

BDD Specs

Manifest

Handover

Product Configuration

ProductTiered Savings Account

CurrencyGBP

Tier 10 – £5,000 @ 1.50% AER

Tier 2£5,001 – £25,000 @ 2.75% AER

Tier 3£25,001+ @ 3.50% AER

Day CountACT/365

AccrualDaily, paid monthly

Interest Simulation

Accrued Interest (£)

Day 1Day 180

DayBalanceDaily Int.Accrued

Day 1£10,000.00£0.48£0.48

Day 30£10,000.00£0.48£14.38

Day 60£30,000.00£1.85£43.26

Day 90£30,000.00£1.85£98.76

Day 180£30,000.00£1.85£265.14

Generated BDD Specification

Gherkin

Feature: Tiered Savings Interest Accrual

Scenario: Daily accrual across tier boundaries

Given: a Vault smart contract "tiered_savings"

And: the customer deposits £30,000

When: accrual runs for 30 days

Then: tier 1 accrues at 1.50% on £5,000

And: tier 2 accrues at 2.75% on £20,000

And: tier 3 accrues at 3.50% on £5,000

Then: total interest equals £55.38 (±£0.01)

Project Manifest

─tiered_savings.pySmart Contract

─test_tiered_savings.pyBDD Tests

─template_params.pyParameters

─.github/workflows/ci.ymlCI/CD

─agents/vault-dev.mdAI Agent

─README.mdDocumentation

↓

Ready for Handover

Tiered Savings Account · 6 files · BDD specs · CI/CD · AI agents

Download ZIP

Open in Lightspeed

Overpass Apex

AI-native banking experiences, governed and production-ready.

The premium tier of Overpass for regulated banking environments. Describe a dashboard in plain language and get working components. Define exactly what your agents can and cannot do. Embed intelligent, context-aware chat into any app. Every AI action is auditable.

For banks ready to put AI in front of users. Controlled, auditable, production-grade.

Stage: Autonomy

Describe interfaces in plain language, get working components

Define, scope, audit, and version agent capabilities

Embed context-aware, KB-powered chat into any app

Manage and distribute knowledge bases visually

Use a banking component catalogue for AI-generated views

Audit every AI action with a complete trail

Overpass Apex

Premium

Gen UI

Tools

AI Chat

Generative UI

Describe the interface you need

Available components

DataTableBarChartStatCardTimelineFormBadge

Generated Dashboard

Live Preview

Overdue

Pending

147

Compliant

1,842

Reviews by Month

XYBC RetailCompliantLow

XYBC WealthPendingMedium

XYBC CardsOverdueHigh

3 components · 1 API call · fully audited

Tool Call Studio

3 tools registered

get_account_balanceread

Fields:balancecurrencystatus

initiate_transferwrite

Fields:amountdestinationreference

modify_interest_rateadmin

Fields:rateeffective_date

Apex ChatKB-powered

What's the current balance on account DE89...4217?

The current balance is €47,230.18 (Active).

viaget_account_balance

Transfer €5,000 to GB29...9268 ref INV-0847

Transfer initiated: €5,000.00 to GB29...9268. Reference: INV-0847. Pending approval.

viainitiate_transfer

Ask anything...

GuardianComing Soon

Deterministic boundaries for AI agents in banking.

A runtime proxy between AI agents and your banking infrastructure. Every tool call is validated against typed schemas, checked against permissions, and filtered for PII before execution. Unregistered tools do not exist. Every decision, whether allowed or denied, is logged in a structured trail you can query and investigate. Prompt guardrails are probabilistic. Guardian enforces boundaries mechanically.

For banks deploying agents in production. Deterministic boundaries between AI and your systems.

Stage: Autonomy

Validate every agent tool call against typed schemas

Check permissions with a JWT-based access model

Filter PII across three tiers: allow, mask, or redact

Block unregistered tools by default. They do not exist

Query every call and every denial in a structured audit trail

Test agent boundaries with an adversarial eval harness

Guardian Runtime

Enforcing

Schema

Permission

PII Filter

Audit

get_account_detailsfrom@cb-devtc-001

Schema—

Permission—

PII Filter—

Audit—

FILTERED·get_account_details→ audit.jsonl

modify_credit_limitfrom@cb-risktc-002

Schema—

Permission—

PII Filter—

Audit—

BLOCKED·modify_credit_limit→ audit.jsonl

execute_shell_commandfrom@cb-reviewertc-003

Schema—

Permission—

PII Filter—

Audit—

BLOCKED·execute_shell_command→ audit.jsonl

initiate_transferfrom@cb-devtc-004

Schema—

Permission—

PII Filter—

Audit—

ALLOWED·initiate_transfer→ audit.jsonl

AI agents are software that can take actions on your behalf, like looking up an account balance or initiating a payment. When these agents operate inside a bank, every action needs to be checked, controlled, and recorded. That's what Guardian does.

Is this a real action?

Every request is validated against a strict definition. If the action isn't registered, it doesn't exist. No exceptions.

Are they allowed to do this?

Each agent has specific permissions. A reporting tool can read data. It can never move money or change settings.

Is sensitive data protected?

Personal information like national insurance numbers and dates of birth are automatically hidden or removed before the agent ever sees them.

Is everything recorded?

Every action, whether approved or denied, is logged with a complete audit trail. Regulators can verify exactly what happened and when.

Most AI safety relies on telling the model to “please don't do bad things.” Guardian doesn't ask. It enforces. Mechanically, every time, with a paper trail.