StaffDeck Open-Sourced: An Enterprise Platform for Building Digital Employees

Introduction

On July 15, 2026, OpenBMB open-sourced StaffDeck, an enterprise platform designed for building, operating, and managing digital employees.

The project takes a different route from the familiar enterprise chatbot. Instead of treating AI as a temporary conversation window, StaffDeck is intended to convert professional experience, business processes, standard operating procedures, and decision criteria into digital employees that can work continuously and improve over time.

The source code is publicly available on GitHub, while desktop packages are offered for macOS, Windows, and Linux. StaffDeck can also be deployed from source with an OpenAI-compatible model endpoint.

From Enterprise Chatbots to Digital Employees

Many enterprise AI products still begin and end with a chat interface. A user opens a conversation, asks a question, receives an answer, and then closes the session.

That model is useful for simple assistance, but it has limitations when a company wants AI to participate in repeatable work.

A digital employee needs more than conversational ability. It must understand its position, follow organizational rules, use approved knowledge, call tools, preserve execution records, and support human intervention when necessary.

StaffDeck is built around this broader idea. It represents digital employees with organizational attributes such as:

• Position and role boundaries
• Employee identity
• Capability profiles
• Knowledge resources
• Skills and SOPs
• Tool permissions
• Work and execution records
• Access scope
• Feedback and improvement history

This structure makes the system closer to a digital workforce management platform than a general-purpose chatbot interface.

Turning Organizational Knowledge into Reusable Capability

One of the most difficult problems in enterprise operations is that valuable knowledge is often scattered across many places.

Some information is written in documents. Some is embedded in spreadsheets and process manuals. A large amount remains in the experience of senior employees, including the small judgments that are rarely recorded in formal SOPs.

When experienced employees leave or change roles, part of that knowledge can disappear with them.

StaffDeck is designed to turn several forms of organizational expertise into persistent digital capabilities:

1. Professional experience
2. Business processes
3. Standard operating procedures
4. Decision criteria
5. Documented knowledge
6. Informal operational rules
7. Feedback generated during real work

The goal is not “digital immortality” in a literal sense. It is a practical attempt to make important expertise reusable, traceable, and easier to improve.

A Joint Project Across Industry and Academia

StaffDeck is not presented as the internal product of a single company.

According to the official repository, the platform is jointly developed by:

• ModelBest
• The NEU–ModelBest Data Intelligence Joint Lab
• Tsinghua University Natural Language Processing Laboratory, or THUNLP
• OpenBMB
• AI9Stars

This combination brings together model research, natural-language processing, enterprise deployment experience, and open-source engineering.

The intended audience includes companies and institutions that want to move AI from individual productivity assistance toward reusable organizational capability.

Core StaffDeck Features

The official project describes four major capability areas.

Build and Manage Digital Employees

Users can define digital employees with their own roles, employee IDs, capability profiles, permissions, and work records.

Digital employees can be published, reused, or adapted for different organizational needs. Permission isolation helps prevent every user from modifying shared resources directly.

This is useful when a company wants to maintain a controlled set of approved digital roles rather than allowing each employee to build an isolated personal bot.

State-Machine-Driven Procedural Skills

StaffDeck can turn natural-language process descriptions into structured SOPs and execute them through state machines.

A state machine helps break complex work into defined stages, transitions, conditions, and branches. This approach is more predictable than asking a model to improvise the entire process inside one prompt.

The platform supports:

• Multiple process flows
• Real-time flow switching
• Context preservation
• Visual process editing
• Version management
• Branch evolution

These features are especially relevant for tasks that must follow a stable procedure but still need limited AI judgment.

Document-Structure-Aware Knowledge Retrieval

Traditional retrieval systems often divide documents into small text chunks and search them directly. This can work, but it may lose the structure of the original document.

StaffDeck builds navigable indexes across several levels, including documents, chapters, pages, and summaries.

A digital employee can first estimate where the relevant information is likely to appear, then locate the original text more precisely.

The platform also supports:

• Separate knowledge buckets
• Targeted retrieval
• Source citations
• Retrieval debugging
• Permission-aware access

This gives teams more visibility into why a digital employee selected a particular source.

Autonomous Execution and Continuous Improvement

StaffDeck can perform work through HTTP APIs, Model Context Protocol tools, and scheduled tasks.

Execution is not treated as a hidden black box. The platform records intermediate events such as intent recognition, retrieval, skill selection, tool use, review, and response generation.

Long-term memory, user feedback, conversation records, and feedback analysis can then be used to improve the employee’s capabilities.

Human users can also intervene, cancel a run, continue queued work, or take over a task when required.

How StaffDeck Preserves Knowledge Beyond a Single Conversation

The most important difference between a digital employee platform and a standard chatbot is persistence.

In a normal chat product, much of the useful context belongs to one conversation. Once that session ends, the knowledge may not become a governed organizational asset.

StaffDeck separates reusable capability from temporary conversation context.

A digital employee can retain and reuse:

• Approved knowledge bases
• Process definitions
• Skills
• Connected tools
• Position-specific rules
• Long-term memory
• Feedback records
• Execution traces
• Scheduled responsibilities

This does not mean every conversation should automatically become permanent memory. Enterprise systems still need permission rules, privacy protection, review processes, and clear retention policies.

The useful part is that knowledge can be deliberately converted into a controlled resource instead of remaining trapped inside an employee’s personal chat history.

Core Workflow

The official StaffDeck workflow can be summarized in six stages.

1. Create a Digital Employee

Define the position, role boundaries, service style, creator, and access scope.

A clear role definition helps prevent the digital employee from acting outside its intended responsibility.

2. Configure Capabilities

Attach knowledge bases, general skills, SOPs, and tools.

Teams can copy resources from the marketplace or create their own without modifying the original shared templates.

3. Start a Session

Open the digital employee from the marketplace or employee list and send the first request.

The formal session is persisted after the first message is submitted.

4. Execute and Observe

Follow the execution record while the digital employee works.

The interface can display events related to intent, retrieval, skills, tools, review, and the final response.

5. Intervene When Necessary

Users can manage queued requests, cancel a run, transfer work to a person, or process pending answers.

Human intervention remains important for uncertain, sensitive, or high-impact tasks.

6. Improve the Employee Over Time

Use memory, feedback, conversation logs, and scheduled tasks to refine the employee’s performance.

The employee becomes more useful when teams review failures and update its knowledge, process definitions, or tools.

StaffDeck Quick Start

The source article announced the open-source release but did not include installation instructions. The following commands are taken from the official StaffDeck repository so readers can verify and test the platform directly.

Requirements

Prepare the following environment:

• macOS, Linux, or WSL for the development scripts
• Python 3.11 or later
• Node.js 20 or later
• npm
• An OpenAI-compatible Chat Completions endpoint
• A valid API key for the selected model service

StaffDeck itself does not require CUDA. Hardware requirements depend on the model endpoint you choose to run or access.

1. Clone and Install

git clone https://github.com/OpenBMB/StaffDeck.git
cd StaffDeck

python3 -m venv backend/.venv
backend/.venv/bin/python -m pip install -e "backend[dev]"
npm --prefix frontend-enterprise ci
cp backend/.env.example backend/.env

2. Configure a Model

Open backend/.env and set the application secret and model-service information:

APP_SECRET="replace-with-a-long-random-secret"
DEMO_MODEL_BASE_URL="https://your-openai-compatible-endpoint/v1"
DEMO_MODEL_NAME="your-model-name"
DEMO_MODEL_API_KEY="your-api-key"

The API key is used to create the initial model configuration and is encrypted before being stored in the database.

Do not commit backend/.env to a public repository.

After the service starts, model providers can also be managed from:

Admin → Model Configuration

3. Launch the Web Demo

DETACH=1 scripts/dev_up.sh

The script builds the frontend and serves the interface, API, and Swagger documentation from a FastAPI process on port 5173.

The initial administrator credentials are:

Username: admin
Password: admin

Change the default password immediately after the first login.

4. Verify the Installation

Run the health check:

curl http://127.0.0.1:5173/api/health

Expected response:

{"status":"ok"}

Then open:

http://127.0.0.1:5173/workspace/gallery

Select a digital employee and send a message. The answer and execution record should appear within the same conversation turn.

Useful Commands

scripts/dev_status.sh       # inspect service status
scripts/dev_down.sh         # stop the local service
scripts/dev_up.sh           # run in the foreground

Desktop Downloads

Users who do not want to deploy from source can use the desktop releases published through the official StaffDeck project.

Platform	Architecture	Official Package
macOS	Apple Silicon, arm64	Download .dmg
Windows	x64	Download installer .exe
Linux	x86_64, Debian or Ubuntu	Download .deb

Because the project is still at an early beta stage, teams should test releases in a controlled environment before using them for important production work.

Project Structure

The official repository separates the application into backend, frontend, documentation, scripts, and packaging components.

StaffDeck/
├── backend/                  # FastAPI APIs, agent runtime, storage, and task workers
├── frontend-enterprise/      # React/TypeScript StaffDeck workspace
├── docs/                     # Tutorials, APIs, schemas, and example flows
├── scripts/                  # Service lifecycle and validation scripts
├── packaging/                # macOS, Linux, and Windows packaging assets
├── README.md                 # English documentation
└── README.zh.md              # Simplified Chinese documentation

The backend is primarily Python-based, while the enterprise frontend uses React and TypeScript.

Enterprise Use Cases

StaffDeck can serve as a foundation for several types of internal digital roles.

Internal Knowledge Assistant

A digital employee can search approved company documents, cite sources, and help employees find process or policy information.

SOP Execution Assistant

Teams can convert recurring procedures into structured flows and allow a digital employee to guide or perform each stage.

Operations Coordinator

With scheduled tasks and connected APIs, a digital employee can monitor routine work and trigger approved actions.

Customer-Service Support

A digital employee can combine product knowledge, service rules, and escalation procedures while preserving execution records for review.

Research or Analysis Assistant

The system can assemble sources, follow a defined research process, and preserve the steps used to produce an answer.

Expert-Knowledge Preservation

Experienced employees can convert recurring judgments, review criteria, and working methods into reusable digital capabilities before changing roles or leaving the organization.

Risks and Limitations

Open-source availability does not make an AI system automatically ready for unrestricted enterprise deployment.

The StaffDeck repository lists several important limitations:

• Model responses may be wrong, incomplete, or inconsistent.
• Execution records improve auditability but do not guarantee correctness.
• Retrieval quality depends on document quality, parsing, indexing, permissions, and model capability.
• External tools can produce real side effects.
• Scheduled tasks require continuously running workers and correct time-zone configuration.
• High-risk actions should use least-privilege credentials and human approval.
• The platform is not a replacement for qualified professional review in regulated fields.
• Important decisions require authorization, privacy protection, and human oversight.

Organizations should begin with low-risk workflows, restrict tool permissions, and review outputs before expanding the system’s responsibilities.

Open-Source License

StaffDeck is released under the GNU Affero General Public License v3.0, commonly known as AGPL-3.0.

Teams planning to modify the platform or offer it over a network should review the license carefully. AGPL-3.0 can create source-disclosure obligations for modified versions made available to users through a network service.

For commercial or legal decisions, organizations should consult the full license text and obtain professional advice where necessary.

常见问题

What is StaffDeck?

StaffDeck is an open-source enterprise platform for building and managing digital employees. It combines role definitions, knowledge, SOPs, tools, permissions, memory, and execution records in one system.

How is StaffDeck different from a chatbot?

A chatbot mainly handles conversations. StaffDeck is designed to support persistent roles, governed knowledge, structured processes, connected tools, scheduled work, audit records, and continuous improvement.

Can StaffDeck run without a local GPU?

Yes. The application connects to an OpenAI-compatible model endpoint, so StaffDeck itself does not require a local GPU. Hardware needs depend on whether the selected model service is hosted locally or remotely.

Which model providers can StaffDeck use?

The platform requires an OpenAI-compatible Chat Completions endpoint. Compatibility depends on whether the provider implements the required API behavior and whether the selected model works correctly with StaffDeck’s workflows.

Can StaffDeck be deployed on Windows?

The official project provides a Windows x64 desktop installer. For source-based development workflows, the repository lists macOS, Linux, or Windows Subsystem for Linux as supported environments.

Why can regular users use marketplace resources but not edit them?

Marketplace resources are shared templates protected by creator and administrator permissions. Regular users can copy or bind authorized resources to their own digital employees without changing the originals.

Is StaffDeck ready for production use?

The project is open source and already includes enterprise-oriented capabilities, but the current release is still an early beta. Organizations should conduct security, privacy, reliability, and workflow testing before production deployment.

What license does StaffDeck use?

StaffDeck uses AGPL-3.0. Organizations modifying or operating the software as a network service should review the license terms and any related source-code obligations.

相关工具

• StaffDeck: The official website for the enterprise digital employee platform.
• OpenBMB: The open-source large-model ecosystem that incubates StaffDeck.
• FastAPI: The Python web framework used to serve StaffDeck’s API and application backend.
• React: The frontend library used in the StaffDeck enterprise workspace.
• Model Context Protocol: A protocol StaffDeck can use to connect digital employees with external tools.
• GitHub: The platform hosting StaffDeck’s source code, issues, releases, and project history.

Related Links

• StaffDeck GitHub Repository: The official source-code repository and primary technical reference.
• StaffDeck Official Website: The official product website and documentation entry point.
• StaffDeck Releases: Official beta releases and desktop installation packages.
• StaffDeck English README: Official features, deployment steps, workflows, risks, and license information.
• StaffDeck Chinese README: Simplified Chinese documentation maintained in the project repository.
• StaffDeck Issues: The official channel for reproducible bug reports and feature requests.
• OpenBMB GitHub Organization: Other models, systems, and open-source projects maintained by OpenBMB.

Summary

StaffDeck represents a shift from temporary enterprise chatbots toward persistent digital employees. It is designed to turn professional knowledge, SOPs, decision rules, tools, and feedback into governed capabilities that can be reused across an organization.

The platform combines role management, state-machine workflows, structure-aware retrieval, API and MCP tool use, scheduled tasks, long-term memory, human intervention, and detailed execution records. Its source code, desktop packages, and deployment instructions are publicly available.

The project is still in an early beta phase, so enterprises should begin with controlled, low-risk scenarios and apply strict permission, privacy, and human-review policies.

StaffDeck’s central idea is straightforward: enterprise AI becomes more valuable when knowledge and processes remain with the organization instead of disappearing when a conversation ends.