Currently in pilot with enterprise teams

Your software runs the business.
Now AI can safely evolve it.

Runtime governs how AI operates inside the legacy systems your organisation depends on. It reduces key-person risk, makes every change explainable, and keeps humans accountable.

Nothing changes without being explained, reviewed, and owned.

Book a walkthrough →Talk to the founder
The reality

Your business depends on software nobody fully understands

The people who built your systems have moved on. The knowledge that keeps things running is undocumented, fragmented, and locked in a few heads. Off-the-shelf AI coding tools make this worse - they add speed without accountability.

Key-person dependency on a shrinking team
Business rules that exist only as tribal knowledge
Pressure to modernise without breaking what works
Board and investor expectations around AI adoption

When software is business-critical, speed without governance is not progress. It's exposure.

Our position

Before AI touches your systems, it must understand them

Most companies try to adopt AI by giving developers faster tools. That optimises the wrong thing. The real bottleneck is not coding speed. It is the fact that system knowledge is fragmented, undocumented, and locked inside individuals. Runtime solves that first.

01AI does not "figure things out." It operates within boundaries your organisation defines.
02Human approval is not optional; it is built into every step.
03Every change has an owner, a rationale, and a paper trail.
What Runtime does

Three problems, one platform

01

Captures what your organisation knows

Runtime consolidates scattered knowledge into a single, persistent, structured record. When someone leaves, the knowledge stays. This is what AI operates against.

{ }
Legacy code
📄
Scattered docs
💬
Tribal knowledge
⚙️
Business rules
Consolidated by Runtime + your team
Canonical Context
ArchitectureConstraintsOwnershipBusiness rulesDependenciesDecisions
Persistent · Queryable · Version-controlled
02

Automates changes safely

AI agents plan, write, and deliver software changes automatically, but only within the boundaries your system knowledge defines. They propose before they act.

Feature request
→
PRD Agent
Requirements
→
Design Agent
ADR
→
Code Agent
Implementation
→
Review Agent
Validation
→
Pull Request
Each agent reads from and writes back to the canonical context
03

Makes every change auditable

Every action has an audit trail. Risk is assessed before execution. Humans approve at defined checkpoints. You can always answer: what changed, who approved it, and why.

👤
Intent reviewed
Human approves scope
⚠️
Risk classified
Impact assessed
👤
Plan approved
Human signs off
⚡
Execution
Within approved scope
👤
Output reviewed
Human validates result
📋
Audit trail
Complete record
Human checkpoint
Risk assessment
AI execution
Audit record

Knowledge reduces risk. Automation reduces dependency. Governance reduces exposure. Together, they let your organisation evolve its software without losing control.

How it works

How Runtime works

From intent to production, with humans in control at every stage.

Human
👥
Stakeholders define intent
Business owners and engineers describe what needs to change and contribute the system knowledge AI depends on.
Change requestsBusiness rulesConstraintsDomain expertise
feeds into
System
🧠
Canonical context
A persistent record of how your systems work. Agents read from this context so they act on facts, not assumptions.
ArchitectureOwnershipDependenciesBusiness rulesDecisions
constrains
AI
🤖
AI delivery chain
Specialised agents work in sequence, each producing a specific artifact. Every agent operates within the boundaries the context defines.
PRD
Requirements
Design
ADR
Plan
Task breakdown
Code
Implementation
Test
Validation
Delivery
Pull Request
reviewed by
Human
✓
Reviewers approve at every stage
Technical leads and domain experts review each artifact before the next agent proceeds. They can reject, modify, or escalate.
Requirements reviewDesign reviewCode reviewFinal approval
delivers
Output
🚀
Production change
Approved changes are delivered with a complete audit trail: who requested it, what was decided, who approved each step, and what changed.
Full traceabilityAudit recordRollback path
Context updated - every change enriches the system's canonical knowledge for future work
Human action
AI execution
Knowledge layer
Production output
Example scenarioPE-backed fintech

A PE-backed financial services company needs to update a 12-year-old payment system to support a new banking partner.

01
System knowledge capturedWhat was in people's heads becomes a permanent record
|Payments-core/Context
System Context
Context files that inform agent behaviourCANONICAL
Constraints
runtime/context/constraints.yaml
Settlement routing logic
runtime/context/settlement-routing.md
Acquirer timeout rules
runtime/context/acquirer-timeout.md
ADR 0001 parallel adapter pattern
runtime/context/decisions/ADR-0001-parallel-adapter-pattern.md
ADR 0002 feature flag routing
runtime/context/decisions/ADR-0002-feature-flag-routing.md
Reconciliation scheduling
runtime/context/reconciliation-schedule.md
Coverage Manifest
runtime/context/coverage_manifest.md
02
AI proposes a planExplains its reasoning before touching any code
|<- FeaturesNew acquirer integration
Intent:
Integrate new banking partner (Acquirer B) alongside existing acquirer. Must preserve settlement routing, respect 8s timeout constraint, and support gradual migration via feature flags.
ADR
State: approved - SHA: a3f91bcapproved
1# ADR: Parallel Adapter Pattern for Acquirer B
2
3## Decision
4Implement Acquirer B as a parallel adapter behind a
5feature flag, preserving the existing Acquirer A path.
6
7## Context
8Current settlement routing uses SWIFT (intl) and SEPA
9(EU domestic). Acquirer A enforces 8s auth timeout.
10Retry logic added in 2019 (undocumented until now).
11
12## Consequences
13+ Zero disruption to existing payment flow
14+ Gradual rollout via feature flag percentage
15- Dual-acquirer reconciliation adds complexity
03
Technical lead reviews and approvesHuman judgement required at every checkpoint
|<- FeaturesNew acquirer integration
Review
State: awaiting approval - Reviewer: Thomas L.
Code diff - 4 files changed+247-18
1// src/acquirer/adapter-factory.ts
2+ export class AcquirerBAdapter implements AcquirerPort {
3+ private timeout = 8000; // from context
4+ private retryPolicy = RetryPolicy.fromContext();
5+
6+ async authorise(request: AuthRequest) {
7+ return this.withRetry(() =>
8+ this.client.auth(request, this.timeout)
9+ );
10+ }
11+ }
Blocked until human sign-off - Step 6 of 9
04
Complete audit trailWhen the board asks what changed, the answer is here
|<- FeaturesNew acquirer integration
Audit trail
Complete - All stages passed - Full traceability
S
Deployed to productionToday 15:01Complete
H
PR approved by Thomas L.Today 14:48Human
AI
Security scan passed - no vulnerabilitiesToday 14:32AI
H
Code review approved by Thomas L.Today 14:18Human
AI
Implementation completed by Code AgentToday 11:45AI
H
Plan approved by Sarah M.Yesterday 16:05Human
AI
ADR generated by Design AgentYesterday 14:22AI
H
PRD approved by Sarah M.Yesterday 11:30Human
S
Feature created: New acquirer integration2 days agoOrigin
9 events - 4 human approvals - 3 AI actions - Context updated with 4 new entries

At no point does anyone ask the AI to "just fix it." They ask it to explain the system first, then propose a plan, then execute under supervision.

Who it's for

Built for organisations where software is business-critical

Runtime is designed for companies that depend on legacy systems to operate: where the software is too important to break, too complex to rewrite, and too understaffed to maintain the old way.

⚙️

PE-backed and mid-market companies

Under pressure to modernise and improve margins, with constrained headcount and legacy systems that underpin daily operations.

🏗️

Organisations with key-person risk

Where critical system knowledge lives in the heads of a small team, and losing one person would be a serious operational risk.

📋

Regulated industries

Where every change to production systems must be explainable and auditable, and where compliance is not optional.

Runtime is not for teams chasing developer speed without governance. It is for organisations where every change carries consequences.

Your software is too important
to change without governance

Runtime gives your organisation a way to adopt AI safely: with full visibility, human accountability, and a permanent record of every decision.

Book a walkthrough →Talk to the founder

If your systems cannot explain themselves, they cannot be trusted.

LLM-agnostic. Runs on European models by default. No data leaves your jurisdiction.