Provides web search and code execution capabilities using OpenAI models (gpt-5/o3), with features like model selection, reasoning effort, search context, conversation continuity, and optional code interpreter for data analysis.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@mcp-ai-assistant-iris search latest advancements in quantum computing"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
An MCP (Model Context Protocol) server that provides web search and code execution capabilities using OpenAI models. The iris tool supports model selection (gpt-5/o3) and optional code interpreter for data analysis.
Named after Iris, the Greek goddess of the rainbow and divine messenger, who swiftly carries information between gods and mortals.
Simply install and use the package from the official npm registry:
claude mcp add iris -s user -e OPENAI_API_KEY=your-api-key -- npx @mokemokechicken/mcp-ai-assistant-irisOr configure manually in Claude:
{
"mcpServers": {
"iris": {
"command": "npx",
"args": ["@mokemokechicken/mcp-ai-assistant-iris"],
"env": {
"OPENAI_API_KEY": "your-api-key"
}
}
}
}Related MCP server: ChatGPT WebSearch MCP
Model Selection: Choose between gpt-5 (default) and o3.
Web Search: Advanced web search capabilities with configurable context size
Code Interpreter: Optional code execution for data analysis and visualization
Conversation Continuity: Continue previous conversations using response IDs
Flexible Configuration: Customizable reasoning effort and search context
The iris tool accepts the following parameters:
input (required): Your question or search query
searchContextSize (optional): Search context size - "low", "medium", or "high" (default: "medium")
reasoningEffort (optional): Reasoning effort level - "low", "medium", or "high" (default: "medium")
model (optional): AI model to use - "gpt-5" or "o3" (default: "gpt-5")
useCodeInterpreter (optional): Enable code interpreter for data analysis (default: false)
previous_response_id (optional): Previous OpenAI response ID for conversation continuity
The iris tool supports conversation continuity through the previous_response_id parameter. This allows you to maintain context across multiple tool calls by referencing a previous response.
Each iris tool response includes a Response ID in the format: [Response ID: resp_abc123xyz]
Use this Response ID as the previous_response_id parameter in subsequent calls
The AI will automatically continue the conversation with full context
When you call the iris tool, the response will include:
The main response content
A Response ID at the end in the format: [Response ID: {response_id}]
First call:
- input: "Tell me about machine learning"
- Response: "Machine learning is... [Response ID: resp_abc123xyz]"
Second call (continuing the conversation):
- input: "Can you give me some practical examples?"
- previous_response_id: "resp_abc123xyz"
- Response: "Based on our previous discussion about machine learning... [Response ID: resp_def456uvw]"Validity Period: Response IDs are valid for 30 days from creation
Context Inheritance: Previous conversation history, tool calls, and reasoning are preserved
Cost Impact: Previous conversation tokens are included in the input token count
Instructions: System instructions are not automatically inherited and must be specified each time
OPENAI_API_KEY: Required OpenAI API key
This project is licensed under the MIT License - see the LICENSE file for details.
iris (v0.6.1): An AI agent with advanced web search and code execution capabilities. Supports model selection (gpt-5.1/gpt-5/o3) and optional code interpreter for data analysis. Useful for finding latest information, troubleshooting errors, and executing code. Supports natural language queries.
| Name | Required | Description | Default |
|---|---|---|---|
| input | Yes | Ask questions, search for information, or consult about complex problems in English. | |
| model | No | AI model to use (gpt-5.1/gpt-5/o3). Defaults to gpt-5.1 (is better). | |
| reasoningEffort | No | Reasoning effort level (low/medium/high). Defaults to medium. | |
| searchContextSize | No | Search context size for web search (low/medium/high). Defaults to medium. | |
| useCodeInterpreter | No | Enable code interpreter for data analysis and code execution. Defaults to false. | |
| previous_response_id | No | Previous OpenAI response ID for conversation continuity. Valid for 30 days from creation. Enables context preservation across multiple tool calls. Use the Response ID from previous iris tool response. |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the full burden. It discloses real behavioral surface — web search, optional code interpreter, model choice, natural language input — but says nothing about cost, latency, network dependence, non-determinism, or the fact that this spawns a full external LLM agent. Those are exactly the traits an agent needs before invoking it.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Front-loaded and compact at four sentences, with capability stated first and supporting details after. The version tag '(v0.6.1)' and the closing 'Supports natural language queries' are minor filler that adds no decision value.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Six parameters, no output schema, no annotations. The description covers capabilities but omits the conversational-continuity behavior (previous_response_id) and any resource/cost expectations, which matter for a 6-param agent tool. It is adequate but leaves an agent without key operational context.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 100% and three parameters carry enums with defaults, so the schema already does the semantic work. The description restates model options (gpt-5.1/gpt-5/o3) without adding format or selection guidance beyond the schema. Baseline 3 is appropriate.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description names specific capabilities — an AI agent doing web search and code execution with model selection and an optional code interpreter. That is a concrete verb+resource picture, not a tautology of the opaque name 'iris'. The only gap is the absence of siblings to differentiate against (there are none), so no routing language is needed.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
It lists use cases (latest information, troubleshooting errors, executing code), which implies when to reach for it, but gives no explicit when-not-to-use or alternatives. With no sibling tools present, the missing routing guidance is a smaller gap, but the 'when should I not call this' side is entirely unstated.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
v0.6.2irisScored across 1 tool
There is only one tool, so there is no possibility of selecting the wrong tool or confusing overlapping purposes. The single 'iris' entry point is unambiguous by construction, though it is broad by nature.
With a single tool named 'iris', there is no inconsistency to detect and the name is short and memorable. However, it follows no verb_noun or other recognizable convention, so a predictable pattern cannot be established.
A single tool for an entire server is very thin; all functionality (web search, code execution, model selection, data analysis) is collapsed into one monolithic call. This leaves no granular surface for an agent to compose or target specific capabilities.
As a general-purpose agent wrapper, the tool can cover a wide range of tasks via natural language, which partially compensates for its thinness. But there is no coverage of auxiliary lifecycle operations (sessions, memory, configuration) that an assistant server would typically expose.
Real-time web search, reasoning, and research through Perplexity's API
Web search, fetch, extract, and research for AI agents. Markdown output + AI-synthesized answers.
Web search, browser automation, scraping, crawling and CAPTCHA solving for AI agents.
Provides AI assistants with access to Seltz's powerful Web Search capabilities.
