Jiki

v1.0.0Developer Toolsstable

MCP server for jiki

grpogrpotrainerinterfacemcpmcp-client
Share:
17
Stars
0
Downloads
0
Weekly
0/5

What is Jiki?

Jiki is a Model Context Protocol (MCP) server that allows AI assistants like Claude, Cursor, and VS Code to mcp server for jiki

MCP server for jiki

This server falls under the Developer Tools category on MCPgee, the world's largest MCP server directory with 33,000+ servers.

Features

  • MCP server for jiki

Use Cases

Integrate GRPO Trainer interface with MCP.
teilomillet

Maintainer

LicenseApache-2.0
Languagepython
Versionv1.0.0
UpdatedAug 2, 2025
Statushealthy
Maintenanceactive

Works with

ClaudeOpenAIwindowsmacoslinux

Installation

Manual Installation

npx jiki

Configuration

Configuration Details

Config File

claude_desktop_config.json

Performance

Response Metrics

Response Time< 200ms
ThroughputMedium

Resource Usage

Memory UsageLow
CPU UsageLow

How to Set Up and Use Jiki

Jiki is a Python framework that acts as an MCP client and LLM orchestrator, connecting language models (Claude, GPT-4, Gemini, and others via LiteLLM) to MCP tool servers. Rather than being a server you expose to Claude, Jiki is a lightweight agent runtime you run locally: it discovers tools from MCP servers automatically, routes LLM queries to those tools, and returns structured responses with optional interaction tracing. It is designed for developers building GRPO trainer workflows and AI agents that need dynamic tool use over MCP.

Prerequisites

  • Python 3.9 or later with pip or uv
  • An API key for your chosen LLM provider: ANTHROPIC_API_KEY for Claude, or OPENAI_API_KEY for GPT models
  • An MCP server script to connect to (Jiki ships an example calculator server)
  • Basic familiarity with MCP server concepts
1

Install Jiki

Install the jiki package from PyPI. Use uv for faster resolution or pip for standard installs.

pip install jiki
# or
uv add jiki
2

Set your LLM provider API key

Jiki uses LiteLLM internally so it works with many providers. Export the key for your chosen provider before running.

export ANTHROPIC_API_KEY=your-anthropic-key
# or for OpenAI:
export OPENAI_API_KEY=your-openai-key
3

Run Jiki in interactive CLI mode

Start the interactive REPL, pointing Jiki at an MCP server script. The --auto-discover flag makes Jiki enumerate available tools automatically.

python -m jiki.cli run --auto-discover --mcp-script-path servers/calculator_server.py
4

Process a single query non-interactively

For scripted use, pass a query directly on the command line and get a structured JSON response with tool call details.

python -m jiki.cli process --auto-discover --mcp-script-path servers/calculator_server.py "What is 42 + 17?"
5

Embed Jiki in your own Python code

Instantiate the Jiki orchestrator in your application to process queries programmatically and access tool call traces.

from jiki import Jiki

orchestrator = Jiki(
    auto_discover_tools=True,
    mcp_script_path="servers/calculator_server.py"
)
result = orchestrator.process("What is 15 * 8?")
print(result)

Jiki Examples

Client configuration

Python code snippet showing how to initialize Jiki with a local MCP server and the Anthropic Claude backend.

{
  "note": "Jiki is a Python MCP client runtime, not a server. Configure it in code:",
  "example": {
    "provider": "anthropic/claude-3-5-sonnet-20241022",
    "mcp_script_path": "servers/calculator_server.py",
    "auto_discover_tools": true,
    "env": {
      "ANTHROPIC_API_KEY": "your-anthropic-key"
    }
  }
}

Prompts to try

Example queries to run through Jiki once connected to an MCP tool server.

- "What is 42 + 17?" (with calculator MCP server)
- "List all the tools available from the connected MCP server."
- "Process this batch of arithmetic problems and return results with full tool traces."
- "What is 100 divided by 4, then multiply the result by 3?"
- "Run the calculator tool to verify: (15 * 8) - (100 / 4)"

Troubleshooting Jiki

Jiki fails with 'AuthenticationError' or 'Invalid API key'

Ensure the correct environment variable is exported for your provider: ANTHROPIC_API_KEY for Claude models, OPENAI_API_KEY for GPT models. Verify the key is valid and has not expired. LiteLLM determines the provider from the model string prefix.

Tool discovery returns no tools from the MCP server script

Verify the path passed to --mcp-script-path is correct and the script is executable. Run the MCP server script directly with 'python servers/calculator_server.py' to confirm it starts without errors. Auto-discovery requires the server to implement the MCP tools/list endpoint.

SSE HTTP transport connection fails when using a remote MCP server

Jiki supports both local stdio (mcp_script_path) and HTTP SSE transport. For HTTP, pass the server URL via the appropriate Jiki constructor argument instead of mcp_script_path. Ensure the remote server is running and accessible from your machine.

Frequently Asked Questions about Jiki

What is Jiki?

Jiki is a Model Context Protocol (MCP) server that mcp server for jiki It connects AI assistants to external tools and data sources through a standardized interface.

How do I install Jiki?

Follow the installation instructions on the Jiki GitHub repository. Clone the repo, install dependencies, and add the server config to your AI client.

Which AI clients work with Jiki?

Jiki works with all major MCP-compatible AI clients including Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Windsurf, and Cline.

Is Jiki free to use?

Yes, Jiki is open source and available under the Apache-2.0 license. You can use it freely in both personal and commercial projects.

Browse More Developer Tools MCP Servers

Explore all developer tools servers available in the MCPgee directory. Each server includes setup guides for Claude, Cursor, and VS Code.

Quick Config Preview

{ "mcpServers": { "jiki": { "command": "npx", "args": ["-y", "jiki"] } } }

Add this to your claude_desktop_config.json or .cursor/mcp.json

Read the full setup guide →

Ready to use Jiki?

Browse our complete directory of 33,000+ MCP servers, read setup guides for your editor, and start building with the Model Context Protocol.

33,000+ ServersFree & Open SourceStep-by-Step Guides