Kanaeru Software Factory
Ship with agents. Stay in control of the outcome.
An opinionated Agent Harness turns product intent into reviewable, evidence-backed software—without letting context, decisions, or accountability disappear between prompts.
10 September 2026 · 00:00 JST
Factory clock
Launches 10 September 2026 at 00:00 JST
Product intro
From a new vision to the next outcome.
One continuous loop, on whichever agent harness and machine you pick: describe a vision, get back outcomes and a plan you can read, see the whole system drawn before a line of code, approve what ships, then follow execution down to the diff and the evidence behind every claim.
Agent speed creates a new coordination problem.
More capable agents do not automatically create a coherent product. The factory keeps intent, handoffs, evidence, and continuity on the same accountable line.
- Intent drifts
- Prompts multiply while the product outcome quietly changes underneath them.
- Handoffs fragment
- Plans, code, reviews, and decisions end up in different tools with no accountable line between them.
- Evidence arrives late
- Teams discover what was actually built only when a change reaches review—or production.
- Context gets rebuilt
- Every new agent, machine, or session spends time reconstructing decisions the team already made.
The Buildline
One governed line from vision to learning
A software artifact advances through six visible stages. Every handoff names the work, the evidence, and the human checkpoint.
Vision
Intent, constraints, and the outcome to pursue
Frame the work and its success conditions
A shared direction with source contextPeople confirm the direction
Plan
Approved direction and repository understanding
Break down work, artifacts, dependencies, and acceptance
A reviewable plan with named readinessPeople approve the plan
Review
Plans, blueprints, gaps, and drift
Stress the proposal before execution starts
Resolved findings, citations, and attestationPeople authorize execution
Execute
Approved tasks and assigned runtime paths
Coordinate agents, machines, and recovery across the work
Task progress, diffs, checks, and visual proofPeople unblock and steer when needed
Deliver
Completed work and collected evidence
Review the shipped change as a governed outcome
Evidence package, delivery review, PRs, and commitsPeople decide what ships
Learn
Delivery record and retrospective feedback
Capture what improved, stalled, and changed
Retrospective metrics and learningsPeople carry the learning forward
One harness turns capability into a production line.
Product intent stays anchored while people approve each stage. The Agent Harness coordinates connected context, governed methods, and supported execution paths against the Buildline—preserving decisions and returning evidence at every checkpoint.
- 50+
- MCP capabilitiesConnected tools made available by role and environment.
- 40+
- Workflow skillsOpinionated methods that shape how work moves through the line.
- 80+
- Specialized agentsFocused workers coordinated around one shared outcome.
- Direction approved
- Execution authorized
- Delivery reviewed
Release inventory: 167 registered MCP tools across 33 families · 46 workflow skills · 88 specialized agents
Join the first release cohort.
Early access is invite-only while we support the first teams closely. Join for the launch briefing; invitations will be sent in batches as capacity opens.