Agent Smith
What is Agent Smith
Agent-Smith is an AI agent designed to work with multiple LLM models, MCP servers, and AI providers that support OpenAI compatible completions API. It serves as an effective tool for Proof of Concept (PoC) and Minimum Viable Product (MVP) development, enabling quick integration testing and model performance evaluation.
Use cases
Use cases for Agent-Smith include developing PoCs for AI applications, testing different LLM models and their performance, and creating customized chatbots for specific tasks or roles.
How to use
To use Agent-Smith, you can run the application by executing ‘go run main.go’. For server mode, use ‘go run main.go --server --port 8008’. The application allows for chat interactions with LLM models, where you can change models, roles, and tools during conversations.
Key features
Key features of Agent-Smith include the ability to chat with LLM models, create customized agent roles, utilize tools from MCP servers, execute Lua code for precise calculations, and manage multiple chats in parallel.
Where to use
Agent-Smith can be utilized in various fields such as software development, AI research, and any domain requiring rapid prototyping of AI integrations or testing of model capabilities.
Clients Supporting MCP
The following are the main client software that supports the Model Context Protocol. Click the link to visit the official website for more information.
Overview
What is Agent Smith
Agent-Smith is an AI agent designed to work with multiple LLM models, MCP servers, and AI providers that support OpenAI compatible completions API. It serves as an effective tool for Proof of Concept (PoC) and Minimum Viable Product (MVP) development, enabling quick integration testing and model performance evaluation.
Use cases
Use cases for Agent-Smith include developing PoCs for AI applications, testing different LLM models and their performance, and creating customized chatbots for specific tasks or roles.
How to use
To use Agent-Smith, you can run the application by executing ‘go run main.go’. For server mode, use ‘go run main.go --server --port 8008’. The application allows for chat interactions with LLM models, where you can change models, roles, and tools during conversations.
Key features
Key features of Agent-Smith include the ability to chat with LLM models, create customized agent roles, utilize tools from MCP servers, execute Lua code for precise calculations, and manage multiple chats in parallel.
Where to use
Agent-Smith can be utilized in various fields such as software development, AI research, and any domain requiring rapid prototyping of AI integrations or testing of model capabilities.
Clients Supporting MCP
The following are the main client software that supports the Model Context Protocol. Click the link to visit the official website for more information.
Content
Description
This is an AI agent that can work with multiple LLM models, MCP servers and AI providers that have OpenAI compatible completions API.
This is great tool for PoC/MVP purposes when you quickly have to play with integrations or test model performance.
Features
- Chat with LLM model. You can change model, role, tools mid-converstaion which allows pretty neat scenarios
- Create customized agent roles via system prompts
- Use tools from MCP servers (both SSE and stdio)
- Builtin tool - Lua code execution when you need model to calculate something precisely
- Multiple chats in parallel
Running
go mod tidy go run main.go
By default it is a desktop application however you can run only the agent part as http server:
go run main.go --server --port 8008
Server API
Server API is at https://unra73d.github.io/agent-smith/
General flow is still designed for desktop app, not for ready-to-go microservice - connect to SSE endpoint to receive updates from server, use other APIs to call features.
Architecture
The diagram above illustrates the internal workflow of the agent, detailing the steps it takes to process requests and generate responses.
TODO
- Chat UI is flaky sometimes
- Calling self as a dynamic agent
- No automatic circuit breakers - user can manually stop generation
- You tell me
Dev Tools Supporting MCP
The following are the main code editors that support the Model Context Protocol. Click the link to visit the official website for more information.
