Langroid Plugin
The langroid plugin gives the agent a curated
library of design patterns for the
Langroid
multi-agent LLM framework. The agent looks up
best practices for agent configuration, tool usage,
task control, and MCP integration — and records
newly discovered patterns for future reference.
Install the plugin via the Claude Code marketplace:
claude plugin install langroid@cctools-pluginsSkills
Section titled “Skills”| Skill | What It Does |
|---|---|
langroid:patterns | Index of Langroid design patterns (agent config, tools, task control, MCP integration) |
langroid:add-pattern | Record new patterns discovered while working with the Langroid codebase |
These skills are meant for the agent to use, not for you to invoke directly. You prompt the agent in natural language and it consults or updates the pattern library as needed.
How to Use
Section titled “How to Use”Finding patterns
Section titled “Finding patterns”Just describe what you want to build and the agent will consult the pattern index:
“Use the proper Langroid pattern to implement an agent that emits a structured ToolMessage and terminates the task.”
“How do I set up batch processing with Langroid so each item gets its own cloned agent?”
The agent uses the langroid:patterns skill behind
the scenes to find the matching pattern and its
code example.
Recording new patterns
Section titled “Recording new patterns”When you discover a useful Langroid idiom while working, ask the agent to record it:
“Record the pattern we just saw for using
@mcp_toolto selectively enable MCP tools on a Langroid agent, using your langroid:add-pattern skill.”
“Add a new Langroid pattern for using quiet_mode() to suppress agent output during batch processing.”
The agent uses the langroid:add-pattern skill to
add an entry to the pattern index and create a
reference document with code examples. Over time
this builds a project-local knowledge base that the
agent can draw on in future sessions.
Pattern categories
Section titled “Pattern categories”The pattern index is organized by category:
- Agent & Task Basics — Setting up agents, wrapping in tasks, returning structured output
- Tool Handlers — Stateful handlers, validation, retry loops
- Task Control — Termination on specific tools, batch processing
- Integration & Output — MCP tool integration, quiet mode for clean CLI output
About Langroid
Section titled “About Langroid”Langroid is a Python framework for building multi-agent LLM applications. It provides:
- A clean agent abstraction with configurable LLMs
- Built-in tool/function-calling support
- Multi-agent task orchestration
- MCP server integration
Visit the Langroid GitHub repository for full documentation and examples.