Open business automation · self-hosted

Run your clients' business on one AI-native platform.

OpenBAM unifies website, marketing, CRM, ERP, e-commerce, workflow and governed AI on one relational core. You deploy it, you operate it — and your clients keep the data.

17 applications · 29 REST API modules · 46 domain plugins and 59 bridge packages on one core.

46+59 domain plugins and bridge packages
17 CGI applications on one core
29 REST API modules, OpenAPI 3.1
118 tables in the relational core
The doctrine

AI proposes. Determinism disposes. Calibration keeps it honest.

The hard problem is not storing data — it is deciding what should happen next in a customer relationship. Rules cannot enumerate the options; an ungoverned model has no authority. OpenBAM orchestrates both, under one governing doctrine.

01

AI proposes

Judgment, ranking, explanation, natural-language interaction. Its output is always a proposal — a suggested transition, a ranked recommendation, a creation plan. AI never writes directly to a business table.

02

Determinism disposes

Workflow guards, hard vetoes, confidence gates and transactional executors keep the last word. A hallucination can waste a query; it cannot corrupt the database or send the wrong campaign.

03

Calibration keeps it honest

Every decision is logged with frozen context, instrument contributions, confidence and outcome. Precision curves and shadow mode decide when probabilistic instruments earn production authority.

Platform

Nine capability domains. One relational core.

Every domain plugin shares the same account, catalog, journey and decision tables — so cross-domain insight costs a join, not an integration project.

Customer journeys

Contact points with signals, context, actions and event monitoring — progression decided by workflow, AI invoked only where judgment is required.

Marketing automation

Universal CTA forms, lead capture and qualification, campaigns, offerings and pipeline, mailing and analytics — all on the same customer record.

CRM and ERP

Accounts, contacts, notes, tasks and documents; a product and service catalog with pricing models, price tiers, usage tracking and fulfillment.

E-commerce

Cart, variants, inventory, shipping, returns, reviews and checkout — layered on the same offering catalog the CRM and ERP already use.

CMS and web presence

Static articles, reusable partials, layouts, YAML menus, media optimization, internationalization and design translation — presentation strictly separated from logic.

BPM and workflow

Visual BPMN modeler, deterministic DSL and AI authoring converge on one Symfony Workflow engine — journeys, forms, agent tasks and decisions share it.

Analytics and BI

Self-hosted web analytics, search, the decision ledger and outcome calibration — the operational core is the intelligence core.

AI services

In-request agents, durable agent runtimes, retrieval-augmented answers, tool calling and governed decisioning — with budgets, approvals and full audit.

Workplace and operations

Social publishing, ads and affiliates, booking, tasks, teams, projects, learning and OIDC single sign-on — one identity across every integrated application.

Decision engine

Every business decision, audited.

The Decision Brain treats each decision as a first-class business object: frozen context, chosen instruments, fused verdict, confidence gate, escalation path and outcome label.

  • Deterministic routing. Which instruments apply to which decision class is a table lookup — never an LLM. The brain obeys its own doctrine.
  • Hard veto and weighted vote. A deterministic veto wins outright; otherwise verdicts fuse by weight times confidence, bounded by a determinism floor.
  • Confidence gate and pragmatism band. Each class encodes the cost of being wrong versus abstaining — ship, escalate or abstain, with the label travelling along.
  • Outcome attribution. Orders, cancellations and journey transitions are attributed back to the decision that preceded them.
  • Shadow mode. Probabilistic instruments earn production authority only after calibration data proves their confidence numbers mean what they claim.
Integrations

An open hub, not a walled garden.

OpenBAM consumes external applications as function-specific containers and exposes itself through a complete REST API, an OpenAPI 3.1 contract, webhooks and MCP.

Mail, marketing and analytics
Listmonk + Amazon SES Umami
Workplace and collaboration
Mattermost HumHub Moodle OpenCloud Google Workspace
Commerce and finance
Medusa InvoiceNinja Lexware
Automation and agents
Activepieces ZeroClaw OBM Harness MCP servers
Search and AI inference
Meilisearch Ollama llama.cpp
Platform and infrastructure
PostgreSQL MariaDB Forgejo WordPress

Identity integration

OpenBAM becomes the single sign-on authority. Integrated applications authenticate through its OIDC provider or delegated OAuth2 flows — one account across the whole workplace.

Data integration

Third-party systems read and write master data, offerings, accounts and orders through OpenAPI-governed REST homes with RFC 7807 errors — never through direct database access.

Automation integration

Deterministic external flows and agent runtimes plug into the decision engine and the workflow engine, and return through the same audited callback path.

For integrators

Build a practice on a platform your clients keep.

OpenBAM gives IT solution providers a sovereign, full-stack platform to implement, operate and extend — with recurring engagement at every layer.

A practice, not a reseller deal

Every deployment carries implementation, migration, integration and operations work. One framework installation serves many client projects, so your platform knowledge compounds across every engagement.

Sovereign by default

Open-source foundations, standard databases, plain-file content and Git-versioned projects. Customers keep their data on infrastructure they control — and you keep control of the stack you operate.

Full-stack surface area

Website, marketing, CRM, ERP, commerce, workflow and governed AI in one platform. One integration project instead of six point solutions — with one data model your clients cannot outgrow.

Git-versioned projects Staging to production chain Container farm with DNS discovery Plugin SDK and Composer packages REST API, OpenAPI 3.1, MCP Decision ledger and audit trails
For SMBs

One platform instead of six subscriptions.

Small and midsize businesses do not suffer from a lack of software — they suffer from too much of it. OpenBAM replaces the silos with one coherent operating layer.

Your integrator deploys it, your team works in it — and the data never leaves infrastructure you control.

  • Change strategy without changing code. Journeys, playbooks, forms and process definitions are data and content — edit, don't develop.
  • See the whole customer. Website behavior, email engagement, pipeline, orders and support context join in one query.
  • Adopt AI without betting the company. Start with suggestions and shadow-mode verdicts; expand authority only as calibration proves each use case.
  • Own the data and the bill. No per-seat, per-contact or per-message licensing; one infrastructure instead of five subscriptions.
  • Software that explains itself. A documentation-grounded chatbot answers in natural language; setup happens in conversation.
How engagements work

From assessment to operating practice.

A repeatable delivery model built on versioned projects and a staged deployment chain.

1

Assess

Map the client's journeys, data domains, integrations and housekeeping requirements. The framework's schema and contact-point model make this a structured inventory, not a discovery workshop without end.

2

Deploy

Stand up the container farm, the framework and the project repositories. OIDC identity, mailer, analytics and backups are part of the installation — staged, versioned and reproducible from Git.

3

Operate and extend

Run the platform: calibration improves decisions, the decision ledger documents them, and new capabilities arrive as plugins or containers — without touching core code.

Trust

Sovereign, auditable, predictable.

The architecture decisions that make OpenBAM safe to operate — and safe to build a business on.

Security in depth

  • bcrypt password hashing, AES-256 and RSA utilities
  • session-appsd encryption keys
  • centralized role-based access control
  • RFC 7807 problem details instead of stack traces
  • container isolation and OIDC/OAuth2 identity

Sovereignty

  • self-hosted deployment end to end
  • credentials in a vault outside the web root
  • push-based, append-only Borg backups
  • plain-file content and Git-versioned projects
  • optional local inference with Ollama and llama.cpp

Auditability

  • decision ledger with frozen context and outcomes
  • append-only agent event logs with replay
  • immutable process definitions, pinned versions
  • approval gates and human escalation paths
  • every publish attempt logged with its result

Cost control

  • per-agent, per-day AI budget caps
  • cheap processors preferred, agents reserved
  • calibration absorbs decisions over time
  • no per-seat, per-contact or per-message fees
  • 100k emails per month at roughly $15
FAQ

Questions integrators ask first.

Is OpenBAM open source?

The foundations are open-source and the platform is self-hostable end to end. OpenBAM ships as a PHP framework plus independently versioned Composer packages — 46 domain plugins and 59 bridge packages at the time of writing — all running in one process against one transactional database.

Can it run fully on-premises?

Yes. The platform runs on Incus containers with unprivileged networking and DNS-based service discovery, standard databases, plain-file content storage and push-based backups. Model inference can stay local through Ollama or llama.cpp — including GPU-accelerated setups — while every model call passes through one provider-agnostic bridge.

How does an IT solution provider earn with OpenBAM?

Through implementation, migration, integration, operations and managed extension work. A single framework installation serves any number of frontend projects, so one platform skill set covers many clients. The container farm and plugin SDK make new capabilities additive rather than invasive.

What does the AI actually cost?

AI cost is governed architecturally: per-agent daily budget caps enforced before each provider request, cheap single-shot processors preferred for ranking, tool-calling agents reserved for investigate-then-decide cases, and locally trained statistical models that progressively absorb decisions. Self-hosted inference converts per-token billing into fixed infrastructure cost.

Does AI write to production data?

No. AI output is always a proposal. Writes go through services and executors that validate the proposal, resolve references and apply it transactionally — or reject it. Workflow guards refuse transitions that do not exist for the current state, and the confidence gate refuses verdicts below the class's threshold.

How do clients migrate from WordPress or existing systems?

Dedicated CMS bridges cover both directions: one imports existing WordPress articles into the OpenBAM CMS, the other serves WordPress as a headless content source rendered through OpenBAM's layout and component pipeline. CRM and ERP data enter through the REST API or the master data concierge.

Get started

Own the stack. Serve your clients.

Create an operator account to explore the platform, or ask the documentation-grounded support chatbot.

Privacy settings

The OpenBAM website uses cookies. Some are essential, while others help us improve the website and your visitor experience.

Statistical data is stored exclusively in anonymized form in Germany and is never forwarded to third parties (such as Google, Microsoft or Facebook).