Your software has a history. Your agents should remember it.

Choose a Software System. Open a session. Bring one repository or many. Konteks catalogues context, estimates cost before work, asks for human approval, ships evidence, and remembers what changed.

Start building Read the launch story

BoundarySystem

MemoryComponent

OutcomeEvidence

konteks://system/acme-checkout● live memory

Software System

Acme Checkout

C4 / catalog System · durable boundary

Files, code, decisions, and history connected as reusable software context

session / 042Reduce payment retries2 Components · proposal readyawaiting approval

session / 038Checkout timeout fix2 repos · PR deliveredreceipt captured

Known Componentsreusable next session

checkout-apigithub.com/acme/checkout-apiverified

payment-workergithub.com/acme/payment-workerlearned

upfront estimate

4–6 SPbefore run ✓

The product model

One System. Many sessions. As many repositories as the outcome needs.

The chat ends. The knowledge does not. Konteks turns session work into organizational memory that the next person or agent can use.

Software SystemCommerce

Session 01Investigate checkout

checkout-api

payment-worker

Session 02Reduce retry noise

payment-worker

event-schema

Session 03“Update checkout”

Known Components resolve naturally.

The launch product loop

Not an agent that merely starts. A system that knows when to stop.

Konteks gives speed a structure: context first, estimate first, human approval, evidence last.

Choose a System

Begin with a durable Software System boundary, not a chat with amnesia.

Ask for an outcome

Speak naturally. Konteks discovers the relevant Components and repositories.

Inspect the proposal

Review architecture impact, tasks, Story Point estimate, and the approval boundary.

Approve the scope

A human binds the exact proposal version and estimate before chargeable work starts.

Agents deliver

Konteks works in isolated, multi-repository workspaces with validation and review.

Receive the evidence

A Value Receipt explains actual cost, outcomes, variance, and newly learned context.

Zero + MCP

People see the product. Agents get the same capability.

Zero is the visual workspace. Core MCP is the door for Codex, Claude, and other MCP-compatible coding agents. Both share the tenant, System, catalog, policy, billing, and evidence—while your team remains free to choose the model.

ZERO / HUMAN

Review with eyes open

Choose System and Components
Compare proposal and estimate
Approve, revise, or cancel
Follow work and open the receipt

Zero is available on the web. iOS and Android apps are coming soon to the App Store and Google Play.

CORE MCP / AGENT

Discover by intent

Ask in natural language
Read the same safe context
Call tools under the same policy
Respect the human approval boundary

The difference that matters

Speed is cheap. Consequences are expensive.

Konteks does not try to be the smartest model. It makes the model you choose informed, bounded, measurable, and accountable enough to work inside a company.

This is not an automation marketplace. Konteks is strongest where your software actually lives: VCS, repositories, the software catalog, Zero, and a public MCP contract for agents such as Codex and Claude.

Memory, not longer prompts

Systems retain Components, repositories, relationships, and session outcomes so the next job does not begin at zero.

Two doors, one product

Zero for people and Core MCP for agents expose the same governed capability.

Estimate before execution

Platform cost is visible before work. Model-provider spend stays separate and goes directly to your provider.

Evidence after execution

Proposal, approval, tasks, commits, pull requests, validation, and settlement form one auditable line.

A team learning to create, review, and share agentic workflows

Agentic enablement

Technology changes the work. People change the company.

We do not believe agentic adoption ends when software is installed. Konteks and Refactory provide agentic training, onboarding, and coaching so your workforce can use agents confidently, safely, and responsibly.

Agentic readiness

Map roles, workflows, risks, and the first valuable use cases before scaling adoption.

Role-based training

Hands-on learning for engineers, tech leads, product leaders, and reviewers working alongside agents.

Guided onboarding

Set up Systems, Components, repositories, policies, and approval patterns for your organization.

Adoption coaching

Build review, measurement, and improvement habits that continue after the workshop ends.

Discuss training & onboarding Available with Konteks adoption or as a tailored engagement.

Honest economics

Pay for the platform that governs the work. Bring your own model.

Every plan is BYOK. Story Points measure Konteks orchestration, execution, validation, and evidence—not disguised provider tokens.

See plans and exact capabilities

Costed Change Proposalreview

Estimated platform work: 4–6 SP
Approved maximum: 6 SP
Model provider: BYOK
Provider estimate: separate
Execution: blocked until approval

A playful software workshop representing a lasting home for repositories and engineering context

31days

00hours

54min

57sec

🇮🇩 17.08

Our first launch note

We thought the problem was the menu. Our software had actually lost its memory.

Read the journey from Refactory’s real-project engineering practice to the System, Component, MCP, estimate, and evidence model behind Konteks.

Read: The road to August 17

One System. A smarter next session.

Stop re-teaching your company to every agent.

Start building