MCP Server Mode
MutagenT provides an MCP (Model Context Protocol) server that allows AI assistants like Claude to interact directly with your prompts, datasets, and evaluations.What is MCP?
MCP (Model Context Protocol) is a standard for AI assistants to interact with external tools and services. When you run MutagenT as an MCP server, AI assistants can:- List and retrieve prompts
- Create and update prompt versions
- Run evaluations
- Start optimization jobs
- View traces and analytics
Installation
The MCP server is a separate package:Claude Desktop Configuration
Add to your Claude Desktop config (~/.claude/claude_desktop_config.json):
Claude Code Configuration
Add to your project’s.claude/settings.json:
~/.claude/settings.json.
Available MCP Tools
Once connected, AI assistants have access to these tools:| Tool | Description |
|---|---|
mutagent_list_prompts | List all prompts in workspace |
mutagent_get_prompt | Get prompt content and metadata |
mutagent_create_prompt | Create a new prompt |
mutagent_update_prompt | Update prompt content |
mutagent_list_datasets | List datasets for a prompt |
mutagent_run_evaluation | Execute an evaluation |
mutagent_start_optimization | Start optimization job |
mutagent_get_optimization_status | Check job progress |
mutagent_list_traces | View execution traces |
Environment Variables
| Variable | Description | Required |
|---|---|---|
MUTAGENT_API_KEY | Your MutagenT API key | Yes |
MUTAGENT_ENDPOINT | API endpoint | No (defaults to production) |
Why MCP Mode?
Context Efficient
Prompt content stays out of AI context window - fetched on demand
Secure
API keys stay local, never exposed in conversation
Real-time
Direct API access, always up-to-date data
Native Experience
AI assistants can naturally work with MutagenT
Example Conversation
With MCP configured, you can have conversations like:You: List my prompts Claude: Uses mutagent_list_prompts You have 5 prompts:
- Customer Support Bot (v2.1)
- Code Review Assistant (v1.0) …
You: Run an evaluation on the Customer Support Bot using the golden dataset Claude: Uses mutagent_run_evaluation Started evaluation run_abc123. Current progress: 45%…