Skip to main content
Version: Next

Obot Chat

Obot Chat is a web-based interface for interacting with AI through customizable projects. Users create projects configured with instructions, knowledge, and MCP server connections, then chat with them through threads.

Projects​

A project combines an LLM, instructions, and connectors to create a specialized assistant. Each project has:

  • Instructions: Define the assistant's behavior and personality
  • Knowledge: Documents uploaded for RAG queries
  • Connectors: MCP servers for external integrations
  • Memory: Important information persisted across all threads
  • Model: The LLM provider and model to use

Creating a Project​

  1. Click New Project in the sidebar
  2. Give your project a name
  3. Add instructions describing what the assistant should do
  4. Configure connectors by enabling MCP servers from the registry
  5. Optionally upload knowledge files

Sharing Projects​

Publish a project as a template for other users to create their own independent copies with their own threads, memory, and MCP server configuration.

Threads​

Threads are individual conversations within a project. Each thread has:

  • Isolated conversation history: Messages don't appear in other threads
  • Independent credentials: Tool authentication is thread-specific
  • Access to shared resources: Knowledge, memory, and project files

Create new threads to start fresh conversations while maintaining the same project configuration.

Tasks​

Tasks automate project interactions through scheduled or on-demand execution.

  • Scheduled: Run on recurring schedules (hourly, daily, weekly)
  • On-demand: Trigger manually or via API
  • Parameterized: Accept inputs to customize behavior

Tasks use the same connectors and knowledge as the parent project.

Creating a Task​

  1. Open a project
  2. Navigate to the Tasks tab
  3. Click New Task
  4. Define the task prompt and any input parameters
  5. Optionally configure a schedule

MCP Server Connections​

Connect to MCP servers through your projects:

  • Browse available servers from the registry
  • Configure credentials per server
  • Enable or disable servers per project
  • All MCP traffic flows through the gateway for access control and logging

Model Providers​

Admins and Owners configure which LLM providers and models are available to users. Users select from these configured models while chatting.

See Model Providers for configuration details.

Knowledge (RAG)​

Add domain-specific information to your conversations:

  • File Upload: Upload documents, PDFs, and text files
  • Smart Retrieval: Automatically finds relevant information for queries
  • Project-Scoped: Knowledge is shared across all threads in a project

Memory​

Maintain important context across conversations:

  • Project Memory: Key information remembered across all threads
  • Automatic Extraction: Important details captured from conversations
  • User Control: Review and manage stored memories

Getting Started​

  1. Create a Project: Define what the assistant should do
  2. Add Connectors: Enable MCP servers for the project
  3. Upload Knowledge (optional): Add documents for context
  4. Start Chatting: Open a thread and interact with your assistant