MCP Resources

Streaming data resources exposed by Nectr MCP server

Overview

In addition to callable tools, Nectr’s MCP server exposes resources — streaming data endpoints that external agents can subscribe to for bulk data access.

Resources use URI patterns and return serialized JSON strings, making them ideal for:

• Bulk data export
• Real-time monitoring dashboards
• Data synchronization with external systems
• AI context augmentation

Resources are read-only. To modify Nectr data, use the REST API endpoints instead.

reviews_resource

Stream recent reviews for a repository as serialized JSON.

URI Pattern

nectr://repos/{owner}/{repo}/reviews

owner · string · required

Repository owner (GitHub username or organization)

repo · string · required

Repository name

Example URIs

nectr://repos/acme/backend/reviews
nectr://repos/openai/gpt-4/reviews
nectr://repos/anthropic/claude-mcp/reviews

Returns

A JSON string containing an array of review objects (same schema as get_recent_reviews tool).

[].id · integer · required

Workflow run ID from the database

[].repo · string · required

Full repository name (reconstructed as owner/repo)

[].pr_number · integer

GitHub pull request number

[].verdict · string · required

AI verdict: APPROVED, CHANGES_REQUESTED, COMMENT, or UNKNOWN

[].summary · string · required

Plain-English review summary

[].created_at · string

ISO 8601 timestamp when the review was created

[].status · string · required

Workflow status: completed, failed, processing, or pending

Example

Request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "resources/read",
  "params": {
    "uri": "nectr://repos/acme/backend/reviews"
  }
}

Parsed Response (Pretty)

When parsed from the text field:

[
  {
    "id": 42,
    "repo": "acme/backend",
    "pr_number": 123,
    "verdict": "APPROVED",
    "summary": "Clean authentication implementation with proper error handling and test coverage.",
    "created_at": "2026-03-10T14:32:15.123456",
    "status": "completed"
  },
  {
    "id": 41,
    "repo": "acme/backend",
    "pr_number": 122,
    "verdict": "CHANGES_REQUESTED",
    "summary": "SQL injection vulnerability in query builder. Missing input validation in user controller.",
    "created_at": "2026-03-09T11:20:03.789012",
    "status": "completed"
  }
]

Implementation

The resource is registered using FastMCP’s @mcp.resource() decorator:

@mcp.resource("nectr://repos/{repo}/reviews")
async def reviews_resource(repo: str) -> str:
    """Recent reviews for a repository serialised as a JSON string."""
    reviews = await _query_db_reviews(repo, limit=20)
    return json.dumps(reviews, indent=2, default=str)

Source code location:

app/mcp/server.py:245

The {repo} parameter in the URI pattern expects the full repository name in the format owner/repo. The FastMCP framework automatically extracts this from the URI path.

Data Limit

The resource returns the 20 most recent reviews for the repository, ordered by created_at DESC.

To fetch more reviews, use the get_recent_reviews tool with a custom limit parameter (up to 50).

Error Handling

If the database query fails:

1. _query_db_reviews() logs a warning
2. Returns an empty list []
3. Resource returns "[]" as the JSON string

The MCP client receives a valid response with an empty array:

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "contents": [
      {
        "uri": "nectr://repos/acme/backend/reviews",
        "mimeType": "application/json",
        "text": "[]"
      }
    ]
  }
}

Use Cases

Real-Time Monitoring Dashboard

Subscribe to the reviews resource and update a dashboard whenever new reviews are posted:

const client = new MCPClient({ url: 'https://nectr.example.com/mcp/sse' });

client.subscribeToResource('nectr://repos/acme/backend/reviews', (data) => {
  const reviews = JSON.parse(data.text);
  updateDashboard(reviews);
});

Data Export

Export all reviews for a repository to a JSON file:

import httpx
import json

async def export_reviews(repo: str):
    payload = {
        "jsonrpc": "2.0",
        "id": 1,
        "method": "resources/read",
        "params": {"uri": f"nectr://repos/{repo}/reviews"}
    }
    
    async with httpx.AsyncClient() as client:
        response = await client.post(
            "https://nectr.example.com/mcp/messages",
            json=payload
        )
        result = response.json()["result"]["contents"][0]["text"]
        reviews = json.loads(result)
        
        with open(f"{repo.replace('/', '_')}_reviews.json", "w") as f:
            json.dump(reviews, f, indent=2)

await export_reviews("acme/backend")

AI Context Augmentation

Provide recent reviews as context to an AI agent:

System: You are a code review assistant. Here are the last 20 reviews for this repository:

<reviews>
{resource:nectr://repos/acme/backend/reviews}
</reviews>

User: What are the most common issues in our codebase?

Slack Bot

Post a summary of recent reviews to Slack every morning:

import slack_sdk

async def post_daily_summary(channel: str):
    reviews = await read_resource("nectr://repos/acme/backend/reviews")
    summary = summarize_reviews(reviews)
    
    slack_client.chat_postMessage(
        channel=channel,
        text=f"Daily Review Summary:\n{summary}"
    )

Comparison: Resources vs Tools

Feature	Resources	Tools
Purpose	Bulk data streaming	Specific queries
URI	Custom scheme (e.g., nectr://)	Tool name (e.g., get_recent_reviews)
Parameters	Embedded in URI path	Passed as arguments
Return Type	Serialized JSON string	Structured dict/list
Limit	Fixed (20 reviews)	Configurable (up to 50)
Best For	Monitoring, export, sync	Single queries, filtering

Use resources when you need to fetch all recent data for a repository. Use tools when you need to query specific PRs or apply filters.

Future Resources (Roadmap)

Planned additional resources:

• nectr://repos/{owner}/{repo}/contributors — Top contributors with stats
• nectr://repos/{owner}/{repo}/health — Repository health metrics
• nectr://repos/{owner}/{repo}/prs/{pr_number} — Single PR review details
• nectr://repos/{owner}/{repo}/memories — Project patterns and memories

Want to request a new resource? Open an issue on GitHub.

MCP Resource Protocol

Nectr follows the MCP resource protocol specification:

1. Registration — Resources are registered with URI patterns during server initialization
2. Discovery — Clients can list available resources via resources/list
3. Subscription — Clients can subscribe to resources for real-time updates (SSE)
4. Reading — Clients can read resource contents via resources/read (JSON-RPC)

List Available Resources

Request

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "resources/list"
}

Authentication

Resources do not require authentication in the current implementation. This is suitable for internal/trusted deployments only.

For production use, add authentication middleware at the FastAPI level.

Source Code

The reviews resource is implemented at:

app/mcp/server.py:245-252

Helper function:

app/mcp/server.py:33-80 (_query_db_reviews)