BotSharp for .NET

A modular .NET framework for multi-agent applications, RAG, and model-provider integration.

Stars
★ 3.1k
Last updated
today
License
Apache-2.0
Primary language
C#

At a glance

Works with
Universal · cross-platformChatGPT · OpenAI API · Claude API
You'll need
.NET 6+API key for the selected LLM providerShell / CLINetwork accessLocal filesystemMCP Server
Typical use
A C# team maintaining an enterprise information system that needs to add agents with distinct responsibilities to an existing workflow.
Main limitation
It requires .NET 6+ and API-key configuration in appsettings.json for the chosen model provider.

What does this agent do, and when should you use it?

BotSharp is an open-source C# and .NET Core framework for connecting language models to existing business applications. Its plugin and pipeline-flow design separates platform concerns into components such as Agent Profile, Conversation & State, Routing & Planning, LLM Provider, caching, file repository, and Rich Content. Built-in plugins cover providers including OpenAI, Anthropic, Google, DeepSeek, Meta, and HuggingFace, along with knowledge-base, Qdrant, channel, and handler extensions. The framework exposes RESTful Open API and WebSocket communication, while a separate SvelteKit project supplies the BotSharp UI. It fits teams that want to self-host and compose multi-agent workflows inside a .NET application rather than adopt a single-provider runtime.

The backend can be started with dotnet run --project ./src/WebStarter/WebStarter.csproj -p SolutionName=BotSharp. Plugin Loader and Hooking load external components, while Agent Profile, Conversation & State, and Routing & Planning organize agent responsibilities, session state, and coordination. Provider plugins such as BotSharp.Plugin.OpenAI, BotSharp.Plugin.AnthropicAI, and BotSharp.Plugin.GoogleAI call supported model platforms. BotSharp.Plugin.KnowledgeBase and BotSharp.Plugin.Qdrant supply RAG-related interfaces and vector search, and BotSharp.OpenAPI plus WebSocket provide application-facing communication. Channel plugins including ChatHub, TelegramBots, WeChat, and Twilio can connect the application to messaging services; built-in MCP integration provides visual MCP management for model-initiated external calls.

  1. A C# team maintaining an enterprise information system that needs to add agents with distinct responsibilities to an existing workflow.
  2. A .NET application team that needs persisted conversation state and routing/planning for coordinated multi-agent tasks.
  3. Developers who need to integrate OpenAI, Anthropic, Gemini, DeepSeek, or HuggingFace through plugins in one application.
  4. An internal-assistant team building retrieval-augmented answers with a knowledge base and Qdrant vector search.
  5. A product team exposing conversational functionality through RESTful Open API, WebSocket, or Telegram, WeChat, and Twilio channels.

How do you install or deploy this agent?

Install the .NET 6+ SDK. Clone the repository, enter its directory, and run:

dotnet run --project ./src/WebStarter/WebStarter.csproj -p SolutionName=BotSharp

On Windows, use:

dotnet run --project .\src\WebStarter\WebStarter.csproj -p SolutionName=BotSharp

For the administration UI, separately clone https://github.com/SciSharp/BotSharp-UI, enter that directory, then run npm install and npm run dev. Before using a model plugin, configure the selected provider's API key in appsettings.json. If dependencies fail at startup, run dotnet restore.

How do you use this agent?

Start the backend and open http://localhost:5015/. Define an Agent Profile and its responsibilities, then select the needed plugins: for example, BotSharp.Plugin.OpenAI or BotSharp.Plugin.AnthropicAI for model access, and BotSharp.Plugin.KnowledgeBase or BotSharp.Plugin.Qdrant for retrieval. Connect a custom frontend through the RESTful Open API or WebSocket, or use channel plugins such as ChatHub, MetaMessenger, TelegramBots, Twilio, or WeChat. If you use the separate administration UI, run npm run dev in the BotSharp-UI project.

What are this agent's strengths and limitations?

Pros
  • C# and .NET Core implementation with plugin and pipeline-flow execution is well aligned with existing .NET business applications.
  • Agent Profile, conversation state, routing, and planning are explicit core modules for multi-agent orchestration.
  • It combines multiple provider plugins with knowledge-base, Qdrant, messaging, RESTful Open API, and WebSocket integration points.
  • Built-in MCP integration includes visual management for connecting model calls to external capabilities.
Limitations
  • It requires .NET 6+ and API-key configuration in appsettings.json for the chosen model provider.
  • The administration UI is maintained in a separate BotSharp-UI repository and adds npm installation and development steps.
  • The supplied material mentions reverse proxies, Nginx/IIS, and separate UI deployment but does not provide concrete production configuration examples.
  • A2A, Computer Use, and Browser Use remain unchecked on the documented roadmap.

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

Agent Source review Stars Updated Language Full support on
BotSharp for .NET This agent 48 · Major gaps ★ 3.1k today C# ChatGPT · OpenAI API · Claude API
Dive into LangGraph 69 · Some gaps ★ 454 12d ago Jupyter Notebook Claude Code
Agentica 63 · Some gaps ★ 352 4d ago Python OpenAI API · Claude API
Hello-Agents 54 · Major gaps ★ 81k 1d ago Python OpenAI API

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Major gaps
48/ 100 5-point scale 2.4 / 5
Trust 7/29
Reliability 8/14
Adaptability 10/18
Convention 11/18
Effectiveness 9/13
Verifiability 3/8
Why each dimension lost points
Trust7 / 29 · 1.2/5

Evidence: The project uses a plugin-based architecture, permissions are distributed across plugins, but no unified permission management mechanism is provided, so least privilege is only partially satisfied (1). User confirmation mechanism is not clearly implemented, such as requiring user confirmation for sensitive operations, so 0. Data flow transparency: README mentions RESTful API and WebSocket, but does not detail data flow and storage, so 1. Sensitive data handling: mentions API key configuration, but no encryption or masking measures, so 1. Dependency security: CI build and tests exist, but no dependency vulnerability scanning, so 1. External effects: plugins can perform external actions (e.g., send messages, call tools), but no restrictions or audit, so 1. Rollback mechanism not mentioned, so 0. Source attribution: project belongs to SciSharp organization, but publisher not verified, so 1.

Reliability8 / 14 · 2.9/5

Evidence: README and code structure are consistent, module division is clear, self-consistency is good (2). Dependency availability: project depends on NuGet packages, CI restores dependencies, but no dependency locking or mirroring, so 2. Failure messages: FAQ mentions common troubleshooting, but no detailed error codes or logging standards, so 1.

Adaptability10 / 18 · 2.8/5

Evidence: Target audience is clearly .NET developers, scenarios cover multi-agent, RAG, etc., so 2. Capability boundaries: core modules and plugins are listed, but unsupported features are not clearly stated, so 2. Trigger precision: routing and planning mechanisms are described, but no specific configuration examples, so 1. Environment fit: cross-platform support, CI matrix, so 2.

Convention11 / 18 · 3.1/5

Evidence: Information architecture is clear, with architecture diagram and module list, so 2. Install notes provide quick start steps, so 2. Naming stability: project name and module naming are consistent, so 2. Examples and FAQ provide multiple examples and common questions, so 2. Known limitations: not explicitly listed, so 1. License: Apache-2.0, complete, so 3. Versioning and changelog: no CHANGELOG provided, so 1. Maintenance responsibility: CI and community support, but publisher not verified, so 2.

Effectiveness9 / 13 · 3.5/5

Evidence: Output usability: RESTful API and WebSocket provided, output format not detailed, so 2. Marginal value: multi-agent framework provides unique value, so 2. Cost-benefit: open source and free, but deployment and configuration costs not assessed, so 2.

Verifiability3 / 8 · 1.9/5

Evidence: Claim traceability: feature claims in README lack specific implementation links, so 1. Cross-source corroboration: CI and documentation exist, but no independent verification, so 1. Fact-inference separation: some content in README is promotional, not clearly distinguished, so 1.

Risks and how to mitigate them
  • Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
  • Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
  • Publisher identity is unverified; assess supply chain risks carefully.
  • No explicit permission management or user confirmation mechanism; strengthen controls for sensitive operations.
  • Dependency security not scanned; perform vulnerability checks before use.
  • Rollback mechanism missing; be cautious with upgrades or changes.
Evidence confidence: Low Reviewed Aug 09, 2026 Reviewed revision e2f09cea0fa7 New commits since this review; the score may not cover them
See the full review method →

FAQ

Is BotSharp free for commercial use?
Yes. It is licensed under Apache 2.0, and the README states that it can be used in personal and commercial projects.
Which model providers are supported?
The README names ChatGPT, Gemini, LLaMA, Claude, DeepSeek, and HuggingFace, and lists built-in plugins including OpenAI, AnthropicAI, GoogleAI, MetaAI, and DeepSeekAI.
Do I have to use BotSharp UI?
No. The framework provides a RESTful Open API and WebSocket communication, so it can be connected directly to a custom frontend or messaging channel.
What should I check if model calls time out?
Check the API key in appsettings.json, network connectivity to the model provider, and whether the configured model name is correct.
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related agents