Grant Workflow MCP Server

Grant Workflow MCP Server

A Python MCP server for querying grant applications and previewing or committing scoped reviewer reassignments.

Category
Visit Server

README

Grant Workflow MCP Server

A small Python MCP server for safely querying grant applications and previewing or committing scoped reviewer reassignments. It was built for the WizeHive take-home exercise.

Stack

  • Python, FastMCP, Pydantic, and PyMongo
  • MongoDB in Docker for local data
  • Anthropic tool-use loop for the optional natural-language agent
  • A JSON golden set and runnable evaluation script

Setup

  1. Create an environment and install dependencies:

    python -m venv .venv
    .\.venv\Scripts\Activate.ps1
    pip install -r requirements.txt
    Copy-Item .env.example .env
    
  2. Start MongoDB and seed the supplied JSON-lines files:

    docker compose up -d
    .\scripts\seed.ps1
    

    The seed script requires the MongoDB Database Tools mongoimport executable. Source data remains in candidate_packet/data/; this avoids maintaining a duplicate copy.

  3. Start the MCP server over stdio:

    python -m server.mcp_server
    

Agent and evaluations

Set ANTHROPIC_API_KEY in .env, then run:

python -m evals.run_evals

The runner prints pass/fail per golden question, aggregate accuracy, per-question latency, and total input/output tokens. It intentionally exits rather than pretending to evaluate if no API key is configured.

Run the lightweight validation tests with pytest.

Tools

The four required tools are search_applications, get_application, cycle_summary, and reassign_reviewer. A very narrow read-only reviewer_workload_summary helper is included because the required agent question ("Which reviewer has the most active assignments?") cannot be answered from the four specified data paths. It prevents the agent from guessing or receiving raw reviewer documents.

reassign_reviewer is preview-first: dry_run=True by default, a cycle or status scope is mandatory, commit requires the exact preview count, and commits over 50 records are rejected. The MCP server never prompts on standard input; confirmation is supplied through the structured expectedCount field so an agent session cannot hang. See DESIGN.md for residual risks.

MCP Server:

Running the MCP server once through MCP Inspector, confirming the five exposed tool names and schemas. <img width="711" height="378" alt="image" src="https://github.com/user-attachments/assets/e6794552-952f-48df-a383-31009ed15f09" /> <img width="705" height="375" alt="image" src="https://github.com/user-attachments/assets/bf9cd3eb-474f-4de1-a985-ebdb77087a39" />

Running safe verification calls to confirm tools:

  1. search_application

<img width="493" height="439" alt="image" src="https://github.com/user-attachments/assets/ed35d6f9-d4db-48bd-b38c-758f433de114" /> <img width="543" height="449" alt="image" src="https://github.com/user-attachments/assets/818cbae8-5ae3-45bb-8eb6-45000781521f" /> <img width="557" height="559" alt="image" src="https://github.com/user-attachments/assets/5e2a87b0-593a-4799-b8e5-ff47da935360" />

  1. get_application <img width="601" height="318" alt="image" src="https://github.com/user-attachments/assets/467f7119-2cfe-4142-b13c-777ccb056f2c" /> <img width="559" height="566" alt="image" src="https://github.com/user-attachments/assets/2b26edb0-39fb-441f-9f62-66c906961232" /> <img width="525" height="557" alt="image" src="https://github.com/user-attachments/assets/49bffcdf-e9d2-408f-a662-4e5dd8a23b97" />

  2. cycle_summary <img width="585" height="330" alt="image" src="https://github.com/user-attachments/assets/924ff834-8cdd-47b3-858e-7f219d2a1398" /> <img width="535" height="412" alt="image" src="https://github.com/user-attachments/assets/63eea109-efb7-46f0-a727-6235c2c58262" />

  3. reviewer_workload_summary <img width="508" height="524" alt="image" src="https://github.com/user-attachments/assets/d01a8df9-178b-4971-b8f2-6fdf38d09e84" />

  4. reassign_reviewer <img width="575" height="426" alt="image" src="https://github.com/user-attachments/assets/4a1c46ef-9475-4c78-a519-c32ee17e03ed" /> <img width="581" height="286" alt="image" src="https://github.com/user-attachments/assets/d0027b80-0f30-4833-a023-6a64cf6603ff" /> <img width="568" height="493" alt="image" src="https://github.com/user-attachments/assets/13d3f031-4b31-414c-bbbe-6249808d9acb" /> <img width="524" height="269" alt="image" src="https://github.com/user-attachments/assets/555c7adc-c9d9-4bc1-9f62-7b5a628727d1" />

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured