Open-source agentic delivery platform

The AI development factory you can run yourself.

AxiaCraft turns product ideas, tickets, incidents, and customer feedback into planned, reviewed, tested, deployable software changes using governed AI agent workflows. It is a product now: open-source, inspectable, and built to live inside your own engineering environment.

Open-source infrastructure for agentic software delivery.

AxiaCraft is a reusable platform for orchestrating AI agents across the software delivery lifecycle. It gives teams the product operations, engineering workspaces, gates, and memory layer needed to run an AI development factory inside their own environment.

Capability 01

Agentic product operations

Automated triage, estimation, prioritisation, ticket shaping, sprint preparation, and delivery status updates.

Capability 02

Autonomous engineering loops

Agent workspaces for implementation, tests, code review, security checks, documentation, and merge-ready output.

Capability 03

Governed release path

Quality gates, reviewer decisions, deployment evidence, rollback notes, and production approval workflows.

A delivery pipeline made of accountable agent work.

AxiaCraft is not a chat window taped onto a backlog. It models the delivery system itself: context, intent, work ownership, verification, approval, and learning.

Step 01Capture signal
Step 02Plan work
Step 03Generate change
Step 04Review and test
Step 05Deploy safely
Step 06Record learning
AxiaCraft delivery timeline

Built around orchestration, evidence, and memory.

The platform gives agents the surrounding system they need to be useful in production: not just prompts, but queues, context, gates, artifacts, and feedback loops.

01

Orchestration layer

Coordinates agents, tickets, approvals, workspaces, retries, and handoffs so the system remains inspectable.

02

Engineering workspaces

Isolated branches and build environments let agents implement and verify changes without trampling active work.

03

Quality gates

Tests, review, security checks, acceptance criteria, and deployment status become first-class product signals.

04

Rembr memory

Durable memory gives agents project context, decisions, runbooks, timelines, contradictions, and causal traces.

05

Human control points

Teams decide where autonomy stops: product approval, architectural risk, production deployment, and incident response.

06

Operational evidence

Every delivery run leaves behind the decisions, checks, artifacts, and outcomes needed for trust and auditability.

AxiaCraft system architecture

For teams that want agentic delivery without losing engineering discipline.

AxiaCraft is aimed at real product teams, internal platform teams, and open-source maintainers who need repeatable software delivery rather than one-off AI demos.

Backlog automation Autonomous fixes Test generation Review workflows Release evidence Runbook learning Agent handoffs Delivery analytics
Why it matters

Agents need a delivery system, not just a bigger prompt.

Useful autonomy comes from boundaries: clear work items, known context, isolated execution, measurable verification, and explicit approval gates. AxiaCraft packages those boundaries into a product so teams can improve the system instead of repeatedly rebuilding the scaffolding.

Run it, inspect it, adapt it.

AxiaCraft is Mark Jones' open-source product for AI-driven software delivery. The goal is a practical foundation that teams can self-host, extend, and reason about without locking their engineering process inside a black box.

axiacraft factory run

input: idea | issue | incident | feedback
plan: classify, estimate, prioritise
build: agent workspace, tests, review
gate: policy, risk, approval, evidence
ship: merge, deploy, announce
learn: memory, timeline, causal trace

status: product-ready platform