Abu Enterprise brings model access, knowledge permissions, policy audit and usage budgets into one private console. Admins can see who uses AI, how much, and what each person can reach.
Employees never handle the real model keys; every call goes through the enterprise's own gateway.
Each department has its own budget, rate limit, allowed models and resource grants.
Organization, department and personal knowledge bases that Abu can search and cite when it answers.
The console, audit records and governance data are deployed in the enterprise's own environment.
Employees are already using AI, and the organization needs to manage it. Once AI reaches a team, the enterprise faces four management problems: cost, access, knowledge and risk.
Every employee sets up a model key. Who uses AI, how much, and which department is over budget are hard for the enterprise to say.
Model calls all go through the enterprise gateway. Admins set budgets, RPM limits and allowed models per department, and review usage and cost in the console.
Models, knowledge bases, Skills, MCP services and file access live in each person's settings, so the enterprise cannot apply one set of organization rules.
Admins draw the boundaries with organizations, departments, roles and pushed policies. The client checks locally before it calls a tool, a Skill or an MCP service.
Policies, FAQs, product material, project documents and meeting notes sit in files, and AI cannot reliably cite the enterprise's own material.
Enterprise knowledge bases support document upload, parsing, chunking, vector search and full-text search. Abu can cite its sources when it answers, and every query is audited.
Employees install extensions and connect internal systems on their own. Source, permission, credential, version and usage records are hard for the enterprise to manage in one place.
Skills go through upload, scan, review, publishing and signature checks. MCP services go through catalog registration, review, authorization and credential management. When something goes wrong, admins can trace it back.
Enterprise adds a governance layer on top of Abu. Employees use AI in the desktop client, and admins configure organization, models, knowledge, tools and governance policies in a private console.
Employees still start tasks, call tools and use Skills in the Abu desktop client. After binding to the enterprise, the client uses enterprise models, enterprise knowledge bases and approved Skills, and shows each employee their own usage and audit information.
Admins configure departments, roles, models, budgets, knowledge bases, Skills, MCP services, policies and audit in the console. Every use of AI in the organization has a boundary and a record, and can be reviewed afterwards.
Six modules: identity, cost, knowledge, extensions, security and operations. Each one states the problem it solves and what the console actually manages.
The enterprise's organizations, departments, employees, roles and login methods are all managed in one console.
Model keys, providers and spend that were scattered among employees come back to the enterprise side for central governance.
Once enterprise documents are ingested, Abu can search and cite them, and every query leaves an audit record.
Internal capabilities and connections to outside tools can all be published, authorized, signature-checked and traced.
Organization-level and department-level policies are pushed to the client, which blocks a sensitive action or asks for confirmation before it happens.
Admins see the governance outcomes, and employees see which of their own data and usage the enterprise has recorded.
Enterprise starts from a private single-host deployment, with the console deployed separately. The client's enterprise capabilities are delivered through enterprise builds and closed-source modules; private modules never enter the public release.
Both editions share the same core, and Enterprise adds organization-level governance. The open-source edition suits individuals and small teams; Enterprise suits organizations that need central control of cost, access, knowledge, extensions and audit.
A local-first desktop app that uses your own model keys and local configuration. Best for individuals, developers and small-team trials.
Free and open source · Local use · Self-managedEmployees keep using Abu while the enterprise manages people, models, budgets, knowledge, extensions, policies and audit in a private console.
Private deployment · Seat license · PoC availableNote: the enterprise MCP client currently provides catalog browsing and authorized credential access, so an enterprise can manage the available tools and the scope of credentials in one place.
Scan the code to add Shawn, Abu's author, on WeChat with the note “Enterprise”, and arrange a demo or a small PoC: settle the deployment, the login method, the model providers, a knowledge-base sample and the first enterprise Skills.
Abu