Product launch · May 23, 2026

Introducing ASE 1.0

The control plane for governed agentic software engineering.

Self-hosted. Identity-aware. Audit-grade. Built for engineering and platform leaders running coding agents inside regulated organizations.

Today we are releasing ASE 1.0 — Agentic Software Engineering — a self-hosted control plane for defining, launching, supervising, governing, and reporting on multi-agent software engineering work.

ASE is built for the engineering and platform leaders who have watched coding agents move from demo to production over the last eighteen months and asked the obvious next question: how do we run this with the same operational seriousness as the rest of our regulated software estate?

This release is our answer.

The problem we built ASE to solve

Coding agents are now genuinely useful. They plan work, write code, run validations, open pull requests, and increasingly cooperate across long-running tasks. In a regulated environment, that capability creates a new operational gap.

A single agent on a developer’s laptop is a productivity tool. A constellation of agents writing production code on behalf of an organization is something else: a workflow that crosses identity, access, change control, audit, and evidence boundaries. Most teams running agentic coding today are stitching that boundary together with shell scripts, ad-hoc logging, and trust.

That works until an auditor, a regulator, or an incident reviewer asks the questions our customers cannot afford to answer with a shrug:

  • Which version of which prompt, model, and policy ran this change?
  • Who approved the launch, and against which definition?
  • What evidence do we have that the agent did not exceed its scope?
  • How do we reproduce, or revert, what an agent did three weeks ago?
  • Who owns the identity and the credentials the agent acted under?

ASE exists so those questions have crisp, defensible answers.

What’s in 1.0

The first end-to-end slice of the platform.

An operator can define a software system, produce versioned governance artifacts, launch a small agent constellation against a real repository, observe it live, and export an audit-grade record of the run.

Governed definitions

Problem, architecture, compliance, constellation, and launch definitions are versioned artifacts with formal JSON schemas and a lifecycle state machine. Nothing reaches production without a definition you can point at.

Guided design sessions

An LLM-assisted workflow walks operators through problem definition and design elaboration. Model selection and credentials are administered through the central model catalog, not scattered config.

Identity, by Keycloak

OIDC Authorization Code + PKCE, MFA, passkeys, and RBAC mapped from Keycloak roles to ASE permissions. Session-scoped final authority lets a designated account become the auditable decision-maker for a resource.

Constellation launch

A model catalog answers which model an agent should use. A harness catalog answers which executable runs it. Built-in harnesses cover Codex CLI, Claude Code, OpenCode, ASE’s multi-provider adapter, and a parity runner.

Live supervision

SignalR-backed dashboards surface claims, leases, branches, work items, gates, failures, and compliance state as work happens — not after.

Tamper-evident audit

Append-only, chain-hashed audit events are written for every significant action, with evidence bundles exportable per run.

Change control

Versioned definitions, approval workflows, and exception justifications are part of the platform — not a process layered on top.

Self-hosted, first and only

Runs on your infrastructure, against your IdP, your repos, your model endpoints. Docker Compose for local; SQL Server + CLAiR Graph (BogDB) for production evidence.

How ASE fits an agentic stack

ASE is deliberately not a monolith. It sits above a small set of clearly-bounded services, each owning what it is best placed to own:

  • Planning service (default provider: Bo) produces decompositions, validation plans, code intelligence, and handoffs.
  • CLAiR owns live coordination — claims, leases, arbitration, blackboard, gates, and work status.
  • CLAiR Graph powered by BogDB stores durable graph-shaped evidence and query projections.
  • Keycloak is the identity and token authority.
  • Coding agents edit code through their own harnesses.
  • ASE owns the human- and product-facing control plane: definition, governance, launch, supervision, reporting, and visual orchestration.

This boundary matters. It means ASE does not compete with the agent frameworks or coordination layers our customers already run. It governs them.

Who ASE 1.0 is for

ASE 1.0 is designed for engineering organizations where agentic coding is no longer experimental and the cost of an ungoverned run is high. That includes:

  • Regulated industries — financial services, healthcare, public sector, critical infrastructure — where auditable evidence is a precondition for adoption.
  • Platform and developer-infrastructure teams responsible for how the rest of the company uses coding agents safely.
  • Organizations with data-residency or sovereignty requirements that rule out hosted control planes by default.

If your team is running agentic coding through shared scripts and personal credentials today, ASE is the operational layer that lets you scale it without inheriting a compliance problem.

What’s next

The 1.0 milestone closes the foundation, identity, definitions, launch, supervision, and audit-export loop. Near-term work is focused on deeper compliance reporting, richer constellation supervision, and broader first-class harness support.

Talk to us

Agentic software work deserves the same operational seriousness as the rest of your regulated estate.

ASE 1.0 is available to qualified design-partner organizations starting today.

← Back to all news

Cookie Compliance

We use cookies to ensure you get the best experience on our website. By continuing to use our site, you accept our use of cookies, privacy policy and terms of service.