Skip to main content

Press Release — ACM Framework v0.5.0

For Immediate Release
October 06, 2025

DDSE Foundation Announces Agentic Contract Model (ACM) Framework v0.5.0 — A Spec-First Contract Layer and Open Reference Runtime for Agentic Systems​

Dhaka, Bangladesh — The DDSE (Decision Driven Software Engineering) Foundation today announced ACM Framework v0.5.0, the first public release of its open, spec-first contract layer for agentic systems and the accompanying Node.js reference implementation. Version 0.5.0 focuses on contract-complete artifacts, deterministic-style execution, and replayable decision memory, enabling engineering teams to build AI-assisted workflows that are auditable, reproducible, and integrable with existing orchestrators.

“Agents only become enterprise-friendly when their plans, tools, and decisions are first-class artifacts,” said Mahmudur R. Manna, Founder of the DDSE Foundation. “With ACM v0.5, we make those artifacts typed, versioned, and replayable so that teams can reason about agent behavior the same way they reason about software.”

What’s in v0.5.0​

  • Contract-first model (Spec v0.5) — Goals, Context Packets, Plans, Capabilities, Tasks, Tools, Policy and Ledger are treated as versioned artifacts. Plans are validated against capability maps before execution to reject hallucinated steps.
  • Structured planning → deterministic-style runtime — A structured planner emits plan candidates with rationale, prompt digests, and tool-call envelopes. The runtime executes plans with guards, retries/backoff, policy/verification hooks, checkpointing, and append-only decision ledger.
  • Replay bundles — Every run can export a single bundle containing planner outputs, tool envelopes, ledger entries, policy decisions, checkpoints, and task I/O — supporting audits, RCA, and regression testing.
  • Adapters & interoperability — Bridges to LangGraph and Microsoft Agent Framework preserve ACM contracts while executing in external engines. MCP (Model Context Protocol) tools are discoverable as first-class integrations.
  • Developer experience — A high-level @ddse/acm-framework façade wires planner, runtime, nucleus configuration, context providers, and adapters so teams can execute ACM-compliant workflows with one call. Reference demos and an AI Coder TUI illustrate end-to-end planning→execution→replay.

Reference Use Cases (included in repo)​

  • Entitlement Decisioning — Policy-gated decisions with auditable rationale and supervisor notification paths.
  • Knowledge Acceleration — Deterministic retrieval + LLM summarization under context snapshots with acceptance checks.
  • Incident Triage — Branching plans with explicit BRANCH_TAKEN records and SLA policy hooks.
  • Invoice Reconciliation — Idempotent compare/log flows with compensation branches and provenance.
  • Agent Coaching — Transparent reasoning capture (nucleus inferences) with verified storage of coaching notes.

These showcase ACM’s contract layer converting stochastic planning into reproducible execution across common enterprise workflows.

Enterprise Benefits​

  • Capabilities OS for every workflow — ACM gives enterprises a governed contract layer that can front any approved capability—internal systems, SaaS adapters, or third-party providers—through one chat or orchestrator surface, with versioned capability maps and deterministic runtimes keeping every action auditable.
  • Replayable trust signals — Each run exports a portable replay bundle capturing prompts, tool calls, policy verdicts, and checkpoints so auditors, regulators, and customer teams can independently verify outcomes.
  • Interoperable by design — Bridges for LangGraph, Microsoft Agent Framework, and MCP let you reuse existing planners, context servers, or adapters while preserving ACM’s contract guarantees.
  • Operational guardrails, not black boxes — Identity, consent, spend controls, and verification hooks ship out of the box; fairness or risk policies stay under your control with full transparency into every decision.
  • Faster path to production pilots — The reference implementations, demos, and AI Coder TUI shorten enterprise onboarding, helping teams prove value quickly without sacrificing governance or extensibility.

Availability & Quick Start​

ACM Framework v0.5.0 is available today under the MIT License. The reference implementation targets Node.js 18+ in a pnpm monorepo. Quick start:

git clone https://github.com/ddse-foundation/acm.git
cd acm/framework/node
pnpm install
pnpm build

# Run a reference demo with vLLM/Qwen (OpenAI-compatible server on :8001)
pnpm --filter @ddse/acm-examples demo -- \
  --provider vllm \  --model Qwen/Qwen3-Coder-30B-A3B-Instruct-FP8 \  --base-url http://localhost:8001/v1 \  --scenario entitlement

Notes on v0.5​

This is the first public release of the ACM spec and reference runtime. Claims in this announcement describe current capabilities and a clear roadmap toward enterprise-readiness. In particular, ACM v0.5 enables auditability, policy enforcement points, and reproducibility; fairness guarantees depend on the policies and datasets organizations configure.

Call to Action​

  • Explore the repo — Read the spec, architecture notes, and examples; export a replay bundle from a demo run.
  • Integrate — Use adapters to execute ACM plans on your preferred orchestrator or agent engine.
  • Collaborate — Join us on GitHub to contribute adapters, verification grammars, or conformance tests.
  • Pilot — Partner with the DDSE Foundation on regulated-industry pilots and evidence-building benchmarks.

Media Contact
manna.mahmud.bd@gmail.com
DDSE Foundation
Project
github.com/ddse-foundation/acm