LanSphere Overview

LanSphere is an enterprise-grade Agent development platform from Lansenger. Deployed privately in your organization’s own environment, it keeps all data inside your domain and pairs fine-grained member roles with resource-level access control, meeting the data-security and compliance requirements of government and enterprise organizations. On the platform, you can build Workflows, Chatflows, and Agents through visual orchestration, then publish your apps to the App Square, a web link, API access, or a Lansenger AI Assistant.

This article covers the platform’s positioning, the differences between the three app types, its core capabilities, and the interface layout. It is written for developers, admins, and business users new to LanSphere, and helps you build a complete picture before you start building.

Platform Positioning

• Part of Lansenger: LanSphere is deeply integrated with Lansenger. Chatflow and Agent apps can be published as Lansenger AI Assistants in one click and listed in the Lansenger client’s “AI Assistant Marketplace.” The platform comes with a built-in Lansenger MCP toolset, so apps can directly call Lansenger capabilities such as messaging, groups, contacts, calendar, video meetings, and media files.
• Enterprise-grade: Built for organizational development and usage, the platform provides member role management, resource-level permissions, ownership transfer, Project Folder collaboration, Conversation Logs, and a Data Dashboard — a complete management toolset that supports multi-person building and operations.
• Private, standalone deployment: The platform is deployed in your organization’s own environment. The Plugin Store offers officially maintained plugin packages and does not connect to any external community marketplace. App data, Knowledge Base data, and runtime logs stay inside your domain at all times.

Three App Types

LanSphere supports three app types. Click “New Project” in the left menu and choose in the pop-up:

Tip: Data is independent across the three app types, and the Workflow and Agent modules do not share data. If you are not sure which type to choose, see the scenario comparison table in “Quick Start.”

More on each type:

• Workflow: Supports three trigger nodes — Input, Scheduled Trigger, and Webhook Callback. Run on demand, triggered automatically on a recurring schedule, or triggered when an external system calls the callback URL. A Workflow can also be published as a component for other apps to call.
• Chatflow: Supports conversational features such as Opening Message, Suggested Questions, Citation & Attribution, Sensitive Word Review, and file upload, and can connect to a Knowledge Base for knowledge Q&A with cited sources.
• Agent: Provides a complete app-development toolset — model selection and configuration, prompt writing, Knowledge Base association, tool configuration, skill orchestration, file configuration, form configuration, and error handling. A published Agent can also be attached as a node inside a Workflow.

Core Capabilities

Visual Orchestration

Build business flows on a canvas by connecting nodes: nodes are added by clicking only (drag-and-drop is not supported); the node list can be expanded/collapsed and searched by fuzzy name match; the bottom of the canvas offers zoom in/out, undo/redo, and change history. The platform provides a rich set of nodes — Input, Scheduled Trigger, Webhook Callback, LLM, Knowledge Base, Agent, Intent Recognition, Conditional Branch, Loop, Iteration, Code Execution, HTTP Request, and more. Before publishing, a Checklist runs automatically to validate configuration completeness and logic.

Knowledge Base RAG

The Knowledge Base offers two build modes — Simple (standard creation) and Expert (Knowledge Pipeline) — with Vector Indexing and Keyword Indexing, supporting vector, full-text, and hybrid retrieval. After creation, verify retrieval quality with the Retrieval Test. You can also connect a Third-party Knowledge Base via API and use it directly, with no data migration. Knowledge Bases give Chatflows and Agents knowledge Q&A capability, with Citation & Attribution supported in answers.

Plugins & MCP

The Plugin Store offers four categories of officially maintained plugins — Models, Tools, Agent Strategies, and Extensions — with commonly used model and tool plugins built in. You can also create Custom Tools to wrap your own APIs, or connect MCP Services from the top-level “MCP Services” menu.

The built-in Lansenger MCP toolset covers six capability domains:

• Messaging: send private/group/app notification messages, get message history, recall messages;
• Groups: create groups, get group details and members, update group members;
• Contacts: get basic contact info, search people;
• Calendar: get the primary calendar, create/query/update events, get the event list;
• Video Meetings: create meetings, pull meeting lists, get meeting recordings and download links;
• Media Files: upload and download media files, get download paths.

SKILL

A SKILL is a reusable capability unit. The platform comes with official skills built in and supports importing skill files and creating custom skills (three types: Code, Workflow, and MCP). Skills can be attached to an Agent, and the LLM decides when to call them based on user intent.

Collaboration Components

Three kinds of reusable units — Snippets, Workflows, and Agents — are consolidated under “Collaboration Components”: Snippets package standardized, common business logic for Workflows to call; Workflows published via “Publish as Component” and published Agents sync here automatically, ready to be called by any app on the platform — no more building the same thing twice.

Publishing Channels

Apps reach users through multiple channels: publish to the App Square for everyone on the platform; enable a web link or API access in the “Publishing Channels” tab; Chatflows and Agents can also be published as Lansenger AI Assistants to reach Lansenger end users. Channels are independent of each other, and a live WebApp can coexist with the publishing channels.

Interface Tour

After logging in, you land on the platform home page with a function menu on the left:

• Account: At the top of the left menu. Click to open the account pop-up, where you can view account info, link/unlink Lansenger, change your password, and enter System Settings (System Users, Models, API Extensions, Language Settings).
• New Project: Below Account. Click to open the app-type pop-up (Workflow / Chatflow / Agent) — the single entry point for creating apps.
• Top-level function menu:

Note: The “Agent,” “Collaboration Components,” and “MCP Services” entries are only visible to the Developer role and above; Viewers can only use published apps in the App Square.

The app details page has five tabs — “App Development,” “Publishing Channels,” “Conversation Logs,” “Data Dashboard,” and “Settings” — covering orchestration, external publishing, run records, metrics, and app configuration. The app card’s more-actions menu offers Edit Info, Permissions, Export YML, Duplicate, and Delete.

Apps in the App Square come in two usage forms by type:

• Conversational app page (Chatflow/Agent): Includes a conversation history panel (new conversation, switch between past conversations), the chat panel, and an app info drawer. The app’s own developer sees an “Edit App” button that jumps to the App Development page.
• Single-run app page (Workflow): Form input plus result display, with no conversation history sidebar.

Roles at a Glance

The platform draws capability boundaries by role. The four roles in one line each:

• Viewer: Use published apps in the App Square and view tool lists; no creation entries.
• Developer: Create and orchestrate apps and Knowledge Bases; save, publish, run, and access APIs; create Custom Tools, MCP Services, and Snippets.
• Admin: Everything a Developer can do, plus publishing as a Lansenger AI Assistant, publishing as a component, installing and deleting plugins, adding members, and managing model credentials and default models.
• Owner: Everything an Admin can do, plus adding Admins, changing member roles, removing members, and transferring ownership.

For the full capability matrix and member-management operations, see “Members & Permissions.”

Typical Workflow

Building and operating an app on LanSphere usually follows this flow:

1. Create the app: Click “New Project,” choose an app type, then start blank, from a template, or by importing a DSL file.
2. Orchestrate and configure: Complete the canvas orchestration or Agent configuration in the “App Development” tab, connecting Knowledge Bases, plugins, MCP Services, and skills as needed.
3. Debug: Verify with single-node and end-to-end debugging, and review past runs in “Debug History.”
4. Save and publish: Save the draft, then publish to the App Square — or enable a web link, API access, or a Lansenger AI Assistant in the “Publishing Channels” tab.
5. Operate and improve: In the “Conversation Logs” tab, review run records and annotate answer quality; in the “Data Dashboard” tab, track success rate, average response time, Token usage, and other metrics — and keep improving the app.

Next Steps

• Not logged in yet? See “Login & Account” for the three login methods and account-binding rules.
• Ready to build your first app? See “Quick Start” and follow the minimal loop through creation, orchestration, debugging, and publishing.