Easy to deploy, stable to execute, sustainable to operate: StaffDeck 1.0.0-preview is now open source
You finally build an agent, and complex tasks fail to execute; you want digital employees to take on real business, only to find documents and data scattered across enterprise systems… The real difficulty of bringing AI into the enterprise has never been just deploying a usable agent. It's letting digital employees connect smoothly to enterprise context, truly take up their posts, execute reliably, and get better over time in continuous operations.
On October 21, 2026, the enterprise digital employee platform StaffDeck 1.0.0-preview was officially open-sourced. This update is a systematic upgrade across deployment, execution, and operations: the new Harness v3 execution kernel, an entry point spanning web and terminal, an independently buildable open-source release package, and an upgrade mechanism that preserves existing data — forming a complete loop for digital employees from "set up" to "in use" to "managed."
From digital employee creation and management in 0.1.0, to team collaboration, enterprise IM, and open APIs in 0.4.0 and 0.5.0, StaffDeck has been doing the same thing: distilling employees' experience, business processes, and judgment standards into digital employees that can work sustainably. At 1.0.0, this path extends further into daily use and long-term operations, making digital employees true partners that are "easy to deploy, stable to execute, and sustainable to operate."
Modular upgrade: freely combine digital employee capabilities
For a digital employee to complete a job, models, processes, knowledge, tools, and collaboration mechanisms must work together. As business scenarios grow richer, making these capabilities easier to combine, extend, and reuse has become the focus of the execution architecture upgrade.
The new StaffDeck uses Harness v3, organizing the digital employee's runtime capabilities in a layered "large module—small module" structure. Each module does its own job and collaborates through unified interfaces, providing an execution foundation that can be assembled per business need. Plug and play, flexible combination.
Eight modules connect the digital employee's complete workflow:
Digital employee management and assembly: define roles, personas, and models; bind capabilities, processes, and teams, so every employee has clear responsibilities and capability scope.
Processes and SOPs: organize business requirements into executable steps, so employees follow established processes: define steps, advance step by step, and check at each stage.
Business capability center: brings together knowledge bases, skills, tools, memory, and execution environments, so employees can look up materials, call APIs, process files, and deliver results. Every call checks permissions and leaves a record.
Interaction and collaboration: connects human handling and digital employee teams, handing tasks to humans when needed, waiting for replies, then resuming execution.
Channel access and task triggering: users can reach employees via Feishu, DingTalk, WeCom, WeChat, the web, open APIs, and scheduled tasks — multi-channel access, ready anytime.
Conversation and execution engine: coordinates task execution, organizes multi-step reasoning and capability calls, and puts employee configuration and business processes into every round of work.
Runtime governance: retains execution traces, call records, and user feedback, providing the basis for process tracking, exception handling, and continuous improvement.
Unified permissions: runs through resource access and capability calls, so digital employees work within clearly authorized boundaries.
Terminal CLI: digital employees enter developers' daily workflow
The new version adds StaffDeck CLI, letting developers log in to the platform, view digital employees, and choose one to start a conversation directly in the terminal. Follow-up questions, task execution, and result viewing can all be done from the command line.
CLI capabilities also extend to the full workflow before and after tasks:
Manage employees and resources: create digital employees, manage skills, tools, and knowledge bases, and complete SOP validation, publishing, and resource binding.
Track tasks and deliver results: query task status, execution records, and final results, and download generated files locally.
Integrate into automated workflows: structured JSON output lets scripts read execution results and connect digital employees to existing development tools and business processes.
The web and CLI connect to the same StaffDeck service, using the same employee configuration, resources, and permissions. Business users maintain knowledge and SOPs on the web, while developers can call the corresponding employee from the terminal and check task execution.
Asynchronous execution for long tasks keeps business processes moving
In real business, many tasks can't be completed in one Q&A; they often involve multi-stage, long-running operations. The new version adds asynchronous execution for tool calls: after a task is submitted, a task ID is returned first, and the system tracks execution status in the background and saves the final result.
Two execution modes for different business interfaces
StaffDeck background execution: runs ordinary synchronous HTTP requests in the background, with no need to modify external interfaces.
Connecting to external async services: saves the task ID returned by the external service, and obtains results by polling status or receiving authenticated callbacks.
Task progress is queryable, execution results traceable
The system continuously saves task status, and callers can query progress and final results by task ID. When failures occur or results are unclear, the corresponding information is retained for follow-up handling.
SOPs save progress and continue after waiting for results
When a process calls an async tool, it saves an execution checkpoint. If the task succeeds, the SOP carries the result to the next step; if it fails or the result is unclear, it carries the error context back to the current step for handling.
Continuous improvements make digital employees easier to use and manage
The new version also improves existing capabilities around team collaboration, long conversations, human intervention, and version upgrades.
More flexible context strategies: building on existing automatic compression, it exposes configurations for context budget, compression trigger ratio, recent conversation turns to retain, and summary budget, applied uniformly to single-employee sessions and team member tasks.
Smoother human handoff: supports specifying notification channels for handlers, improves identity and channel reachability checks, and supports replying to handoff notifications by quoting Feishu messages, connecting human handling with process resumption.
Clearer team collaboration: improves multi-employee task orchestration, shows the owner's summary progress, and supports viewing the full team log online to understand division of labor and execution.
Start with a simple task to build your own digital employee
For teams using it for the first time, start with a clear, small task: choose a preset employee, configure a model, connect a business tool, and organize the handling steps into an SOP.
Business users maintain knowledge and processes on the web, developers integrate automation via CLI and API, and administrators check task results through execution records. Validated skills and processes can be gradually reused across more digital employees.
StaffDeck 1.0.0-preview hopes to make the value of digital employees show in continuous work: tasks have an execution basis, processes can be inspected, results can be delivered, and accumulated knowledge can carry forward.
Welcome to try it, and to share scenarios, problems, and suggestions from real use with the community.
How to get it
➤ Official website
🔗 https://staffdeck.openbmb.cn/
➤ GitHub repository and downloads
🔗 https://github.com/OpenBMB/StaffDeck
➤ Installation and upgrade guide
🔗 https://github.com/OpenBMB/StaffDeck/blob/main/INSTALL.md
➤ CLI guide
🔗 https://github.com/OpenBMB/StaffDeck/blob/main/CLI.md