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., "@ai2dev-mcp Design a simple note-taking app with offline sync."
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 server that a frontend can talk to in order to design software projects the same way AI2DEV does: draft a design document with an LLM, refine it based on feedback, answer follow-up questions, and finally hand the finished document to the AI2DEV API to create the project.
Tool | What it does |
| Calls the LLM to draft a structured markdown design document from a project name, requirements, and optional audience/constraints. |
| Calls the LLM to revise an existing design document based on feedback, keeping its structure. |
| Calls the LLM to answer any question, optionally grounded in supplied context (e.g. the current design document). |
| Calls the AI2DEV API to create a project from a finalized design document. |
Related MCP server: Lockstep Core
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
cp .env.example .env # fill in ANTHROPIC_API_KEY and AI2DEV_API_KEYRequired environment variables:
ANTHROPIC_API_KEY — used by the LLM tools (design doc generation/refinement, Q&A).
ANTHROPIC_MODEL — defaults to claude-opus-4-8.
AI2DEV_API_BASE_URL — base URL of the AI2DEV API (defaults to a placeholder; set to your real endpoint).
AI2DEV_API_KEY — bearer token for the AI2DEV API.
As a stdio MCP server (what most MCP clients/frontends expect):
python -m ai2dev_mcp.server
# or, after `pip install -e .`
ai2dev-mcp-serverFor local interactive testing with the MCP Inspector:
mcp dev src/ai2dev_mcp/server.pyPoint your MCP client at the command above and pass the environment variables
through its env config, e.g. for a JSON-based MCP client config:
{
"mcpServers": {
"ai2dev-design": {
"command": "ai2dev-mcp-server",
"env": {
"ANTHROPIC_API_KEY": "sk-ant-...",
"AI2DEV_API_BASE_URL": "https://api.ai2dev.example.com",
"AI2DEV_API_KEY": "..."
}
}
}
}pytestsrc/ai2dev_mcp/
config.py # env-driven settings
llm.py # Anthropic-backed design doc generation/refinement/Q&A
ai2dev_client.py # HTTP client for the AI2DEV project-creation API
server.py # FastMCP server wiring the tools together
tests/
test_server.pyAnswer a free-form question, optionally grounded in supplied context (e.g. a design doc).
Args: question: The user's question. context: Optional supporting context, such as a design document.
Inputs: question (required), context. Returns: result
Create a new project in AI2DEV from a finalized design document.
Args: project_name: Name for the new AI2DEV project. design_document: The finalized design document content. description: Optional short project description.
Inputs: project_name (required), design_document (required), description. Returns: result
Generate a software design document from a project description using an LLM.
Args: project_name: Name of the software project. requirements: Free-form description of what the project should do. target_audience: Optional description of who the software is for. constraints: Optional technical or business constraints to respect.
Inputs: project_name (required), requirements (required), constraints, target_audience. Returns: result
Revise an existing design document based on feedback, preserving its structure.
Args: design_document: The current design document (markdown). feedback: Requested changes or additional context to incorporate.
Inputs: feedback (required), design_document (required). Returns: result
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 4 tools
Most tools target distinct phases (design doc generation, refinement, project creation), but 'ask_question' is generic and could be confused with other tools if used for design-related queries, though its purpose is broader.
Three tools follow a clear verb_noun pattern (generate_design_document, refine_design_document, create_ai2dev_project), but 'ask_question' breaks the pattern and does not reference the domain, causing minor inconsistency.
With 4 tools covering the core design-to-project workflow plus a general Q&A, the count is well-scoped and feels neither too sparse nor overwhelming.
The core flow (generate doc, refine doc, create project) is covered, but missing operations like listing, updating, or deleting projects create notable gaps for a complete lifecycle.
MCP server for generating rough-draft project plans from natural-language prompts.
MCP server that lets your AI assistant build and deploy apps, hosting each one at a live URL.
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
Driflyte MCP server which lets AI assistants query topic-specific knowledge from web and GitHub.
