Centralize model access, knowledge permissions, policy audit, and usage budgets so the enterprise keeps cost, data, and risk under control.
The open-source edition helps individuals move faster. At team scale, enterprises need to manage cost, access, knowledge, and risk.
When everyone brings their own model key, it is hard to see usage, spend, and over-budget teams.
Route model calls through the enterprise gateway, set team budgets and RPM limits, and review usage and cost in the console.
Models, knowledge bases, Skills, MCP services, and file access cannot live only in personal settings.
Use organizations, departments, roles, ACLs, and client-side policy checks before tool, Skill, and MCP calls run.
Policies, FAQs, product docs, project files, and meeting notes need to become trusted AI context.
Enterprise KBs ingest documents, chunk and embed them, retrieve with hybrid search, cite sources, and audit queries.
User-installed extensions and internal-system connections need source, permission, credential, version, and usage records.
Skills go through upload, scan, review, publish, and signature checks. MCP services go through catalog, review, ACL, and credential governance.
Employees use AI in the desktop client while administrators manage organization, models, knowledge, tools, and governance policies in a private console.
Employees keep using Abu to start tasks, call tools, and run Skills. After binding, the client uses enterprise models, knowledge bases, approved Skills, and transparency views.
Admins manage departments, roles, models, budgets, knowledge bases, Skills, MCP services, policies, and audit so AI usage has boundaries and records.
These are grouped around the practical enterprise questions buyers ask during rollout: what can be configured, distributed, controlled, and traced.
Bring organizations, departments, employees, roles, and login methods into one control plane.
Centralize scattered model keys, providers, and AI spend in the enterprise control plane.
Turn internal documents into trusted context Abu can retrieve, cite, and audit.
Make internal capabilities and external tool connections publishable, authorized, signed, and traceable.
Push org-level and team-level policies to the desktop client so sensitive actions are blocked or confirmed before execution.
Admins see governance outcomes. Employees can inspect their own enterprise data and usage.
The enterprise console is deployed separately. Enterprise desktop capabilities are delivered through enterprise builds and private modules, not through the public release package.
The standard deployment includes Console, Postgres/pgvector, LiteLLM, Nginx, Redis, MinIO, and optional observability or KB enhancement components.
PoCs and 50-500 person teams can start with the default deployment. 500-2,000 seats should tune database, gateway, and observability resources. Larger environments need a dedicated HA design.
The database layer stays close to standard Postgres, images support amd64/arm64, and local model or private-provider paths can be validated for each customer environment.
The open-source edition is for individuals and small teams. Enterprise is for organizations that need centralized cost, access, knowledge, extension, and audit control.
Local-first desktop app with user-managed model keys and local configuration. Best for individuals, developers, and small-team trials.
Employees keep using Abu while the enterprise controls people, models, budgets, knowledge, extensions, policies, and audit in a private console.
Note: enterprise MCP in the current client focuses on catalog browsing and authorized credential access, so organizations can centralize approved tools and credential scope first.
Start with a demo or small PoC: deployment shape, login method, model providers, knowledge samples, and the first approved enterprise Skills.