Abada Platform documentation

curl -fsSL https://install.abadaplatform.com | bash

The transactional ACID rail

PostgreSQL is authoritative. Lock, validate, advance, persist, write history, facts and outbox records, then commit as one workflow-state transaction.

Studio is the authoring surface

An AI-native single UI: canvas and APL YAML are one document, Dry Run stays local and mocked, and Deploy & Start is the durable engine action.

Agents are durable work

Agent nodes compile to external tasks on the worker protocol. The engine never invokes a model inside a workflow transaction.

Optimization is governed

The Insight Engine turns terminal facts into validated APL proposals that reach production only through explicit human review. No auto-apply.

From BPMN to agentic orchestration

Abada grew up as a BPMN engine and keeps that heritage: BPMN 2.0 files import, validate and run on the same state machine. The platform has since evolved into agentic orchestration on a transactional ACID rail:

• APL is the native authoring language. Studio serializes its canvas to abada.io/v1 YAML, and the engine compiles that YAML directly into the executable graph — no XML round trip. BPMN remains a first-class import and backward-compatibility format.
• Probabilistic work is bounded by deterministic rules. The doctrine is the table is the law, agents are the advice: decision tables evaluate inside the transaction, agent steps run as durable external tasks, and approval gates put a human between them.
• Agentic capabilities ride the same rail. The first-party Agent Worker and the Insight Engine reuse the exact PostgreSQL-authoritative, atomic command machinery that certified the BPMN runtime.

Documentation tracks

This site serves four audiences:

1. User guide — deploy the platform, author your first agentic workflow in native APL, run agent nodes, operate Studio, and review Insight proposals.
2. Architecture — the transactional execution core, the APL canonical model, Agent Worker execution and the Insight Engine, plus deployment, cluster and security boundaries.
3. Compatibility & migration — BPMN import, dialects, the supported subset and the Camunda 7 path.
4. Developer guide — navigating the repository, extending the engine and contracts, and selecting the required evidence.

Release status

These guides describe the platform as it exists on the current dev tree.

• 1.0.0-rc.7 — prepared evaluation release candidate. PostgreSQL is the production source of truth, API v1 and worker protocol v1 are frozen, and the BPMN subset and native APL are certified runtime surfaces. The development installer enables the Agent Worker and the Insight Engine out of the box.
• 1.1 RC agentic track — delivered in the working tree. Durable agent attempt metadata, worker health, the PostgreSQL-authoritative Insight loop and the Studio governance surface are implemented. Production-scale evidence and the cloud infrastructure certification track remain open; no production-certified 1.1 claim is made until they close.

Status callouts throughout this site mark which capability belongs to which tier. The authoritative gates live in the release roadmaps and contracts.