Knowledge Worker Agent
Knowledge Worker Agent for Citrix SecurSpaces™ gives people outside engineering an AI agent that does real work, inside a workspace your organization controls. It reads and writes documents, uses the business tools you connect, and carries out multistep tasks, such as drafting a report from source documents, preparing a customer briefing, or analyzing a spreadsheet. People work with it in a chat, not in a code editor.
Important: Knowledge Worker Agent is an open-source project, released under the Apache License 2.0 on GitHub. It isn't a feature of SecurSpaces, and Citrix doesn't provide support for it. Use it as an example of what you can build on SecurSpaces, and review it before you use it with your own data.
Why use it
Teams such as finance, legal, operations, and sales want AI agents that do more than answer questions. Personal AI subscriptions give each employee a general-purpose assistant, but the organization can't see or shape how it's used. An agent that runs on an employee's laptop can also reach everything that employee can.
Knowledge Worker Agent takes a different approach:
-
Your organization sets the boundaries. The agent runs in a SecurSpaces workspace. The workspace holds the files and tools a task needs, and administrators decide what else it can reach and which models it uses. These are the same guardrails SecurSpaces applies to AI coding agents, so one platform covers agentic work for developers and business teams alike.
-
Teams start with an assistant that knows their work. Administrators provide agents, skills, context, and connectors for each project. A sales team can start with an agent that prepares customer briefings in the house format, and a product team with one that drafts tickets from their backlog. The organization's know-how lives in the configuration, not in each person's own prompts.
-
It works with your data, not only what people paste in. Connectors link the agent to the services your teams already use, and each user signs in with their own account.
-
Work builds up over time. Projects keep related chats, files, and standing instructions together, and the assistant can search earlier chats, so work picks up where it left off.
-
It can work while people are away. Scheduled tasks run a prompt at set times, so a weekly summary or a morning briefing is ready when it's needed.
For what the platform provides, see How Knowledge Worker Agent works.
What you get
| Capability | What it means | Learn more |
|---|---|---|
| Your choice of model | Use a commercial model provider or a model you host yourself, and change it later. | Connect to a model provider, Use a self-hosted model provider |
| Runs where your data is | SecurSpaces runs in your own environment, including air-gapped sites. With a self-hosted model, Knowledge Worker Agent doesn't need internet access. | Where SecurSpaces runs |
| Central management | Administrators assign connectors, agents, skills, and context from one configuration repository. Each user gets them when their workspace starts. | Set up central configuration |
| Connects to your tools | Connectors use the open MCP standard, so the assistant can work with the services your teams already use. | Manage connectors |
| Work that carries over | The assistant can search your earlier chats, so long-running work doesn't start from nothing. It sees only your own history. | How the assistant uses your past chats |
| Projects | Keep related chats, shared files, and standing instructions together. | Organize work in projects |
| Shared reports and dashboards | Share a file you built in a chat with people in your project. They open it in their browser, and always see its latest version. | Share a file |
| Scheduled tasks | Run a prompt on a schedule, and review each run's results. | Schedule tasks |
| Visible AI usage | If you route AI traffic through an AI Gateway, requests from Knowledge Worker Agent are attributed to the user, project, and workspace they came from. | Configure AI Gateway routing |
| Open to inspection | The source is open, so your security team can review what the agent does before you roll it out. | Knowledge Worker Agent on GitHub, How Knowledge Worker Agent works |
Start here
| If you want to… | Go to |
|---|---|
| Try it for the first time | Your first chat |
| Understand where the assistant runs and what it can reach | About |
| Get everyday tasks done in a chat | Use Knowledge Worker Agent |
| Set it up for the workspaces in a project | Administer Knowledge Worker Agent |
| Look up commands, settings, and formats | Reference |
Before you roll it out
By default, the assistant acts without asking. It can change any file the workspace user can and run any command. Scheduled tasks run with nobody there to approve them. Model providers receive the conversation and whatever the assistant reads to answer you. Read How Knowledge Worker Agent works before you give it to other people. For the controls the platform applies to every workspace, see Security and compliance.