Skip to content

README.md

Latest commit

 

History

History
75 lines (51 loc) · 2.68 KB

README.md

File metadata and controls

75 lines (51 loc) · 2.68 KB

AI Agent Context Server

local-context-server is a lightweight, local MCP (Model Context Protocol) server designed to provide long-term memory for AI agents like Gemini, Claude, and others within environments like Cursor.

It allows an AI agent to save, load, list, and search for "contexts"—pieces of knowledge such as application maps, test data, or business requirements—making it possible to create adaptive and intelligent QA automation.

Features

  • Persistent Memory: Store any JSON-serializable data in a local SQLite database.

  • Simple Tool API: Provides four core tools for the AI to manage its knowledge:

    • save_context: Save or update a piece of knowledge.
    • load_context: Retrieve knowledge by its unique ID.
    • list_contexts: Browse all available knowledge.
    • search_contexts: Search for knowledge by keyword.

  • Automatic Database Location: The server automatically creates and manages its database file (memory_tests.db) in a folder named context-database inside your user's home directory. This requires no configuration.

Installation & Usage

The server is designed to be run on-the-fly by an MCP client like Cursor, requiring no manual setup for end-users.

Prerequisites

Python 3.8+ pipx (a tool for running Python applications in isolated environments)

pip install pipx
Usage in Cursor

To use this server with Cursor, add the following to your .cursor/mcp.json configuration file. This command downloads and runs the server package from PyPI.

"mcpServers": {
    "context-local-server": {
      "command": "pipx",
      "args": [
        "run",
        "context-local-server"
      ]
    },
}

How to Use with an AI Agent

Once configured in Cursor, you can interact with the server using the name you defined in mcp.json.

Saving a Context

Your mission is to execute the login flow and document the steps. As you perform each action (navigate, enter text, click) with @mobile-mcp, add a description of that action to a list.

When you have successfully logged in, use @local-context-server to save the complete list of steps with the id flow_login_v1.

Listing all Contexts

Use @local-context-server, execute list_contexts to show me all the knowledge you have.

Searching for a Context

Your mission is to perform the login flow. I don't remember the exact ID. Use @local-context-server to search for contexts related to "login" (search by keyword "login" or related), load the correct one, and then execute the steps using @mobile-mcp.

Using a Context

Perform a login using @mobile-mcp. Use the context app_map_v1 from @local-context-server as your knowledge base for the element IDs.