- Home
- Products and Trials
- Product: ASE
Agentic engineering
with operational discipline.
Your developers are pointing AI agents at production code today. ASE is the system that defines what those agents are allowed to do, supervises what they actually do, and produces the audit trail when someone asks.
For the engineering leaders, executives, and compliance officers responsible for what AI ships into production.
See ASE running against a real repo. No deck.
A handful of agents loose on your codebase is not software engineering.
ASE brings to agentic delivery what regulated workflow automation brought to business operations: identity boundaries, versioned definitions, approvals, immutable audit, evidence reporting, and recovery paths — in one product surface your engineering and compliance teams can both stand behind.
Agents are already touching your repos.
Copilot, Cursor, Claude Code, and homegrown wrappers are writing, refactoring, and merging code in your codebase right now. None of that is captured in your existing change-control system.
No definitions, no approvals, no record.
Prompts are improvised. Reviews are best-effort. Logs are scattered across vendors. There’s no single artifact your auditor, your CFO, or your board can ask for — and no one in the room can answer for what the agents decided last week.
Govern the agents the way you govern people.
ASE captures the work definition, the approvals, the live state, and the evidence trail for every agent constellation on your codebase — with identity, versioning, and tamper-evident audit your existing controls already speak to.
ASE is not a prompt-chain runner. It sits above your code-intelligence providers, your live coordination service, your graph evidence store, and your model-specific coding agents — and gives the engineering lead one product surface for the full lifecycle of an agent constellation working on a real codebase.
What ASE owns
- Project & problem definition — guided, LLM-assisted intake captured as versioned definitions.
- Design elaboration — architecture, interface, and compliance plans, version-tracked.
- Compliance control matrix — mapped to your regulatory regime, with exception/approval workflows.
- Agent constellation configuration — capability inventory, launch profiles, work distribution policy.
- Live orchestration dashboards — claims, leases, gates, branches, work items, failures, compliance state.
- Change control — human approval gates around merges and exceptions.
- Tamper-evident audit — immutable evidence and compliance reports you can hand to auditors.
What ASE delegates
- Identity to Keycloak — OIDC, MFA, passkeys.
- Live claim arbitration to CLAiR — leases, blackboard, gate enforcement.
- Code intelligence to the Planning service (Bo by default) — decompositions, validation plans, ACOP artifacts.
- Durable graph evidence to CLAiR Graph powered by BogDB.
- Model-specific code editing to the agents themselves — Codex, Claude, Gemini, your local workers.
Each authority boundary is a contract, not a black box — you can swap providers without rewriting your control plane.
A realistic first slice
The first useful ASE slice lets an operator stand up a small agent constellation against a single repository — with the controls and evidence you'd expect for production work.
- Define a software system or problem through a guided, LLM-assisted interface.
- Produce versioned problem, compliance, and constellation definitions.
- Connect to your CLAiR Keycloak realm, CLAiR coordination service, Planning provider (Bo), and CLAiR Graph (BogDB) store.
- Launch a small agent constellation against one repository.
- Observe live claims, leases, work state, and failures in real time.
- Export an audit and compliance report for the run.
Built for self-hosting
Why self-hosted
Your source code, your compliance posture, your identity authority. ASE deploys inside your network alongside the rest of your engineering stack — no agentic activity, evidence, or definitions leave your perimeter unless you choose to export them.
For regulated workloads, this is not a nice-to-have. Auditors need to see who approved what, what evidence was captured, and what the agent constellation actually did — with the same trust model you already apply to humans on the same codebase.
Bring agentic delivery into your engineering org — without losing control.
ASE is in active development at Beyond Ordinary. If you’re trying to operationalize coding agents against a real, regulated codebase, the 30-minute walkthrough is the fastest way to see whether ASE fits your stack.