FetchNode Python & Django SDK

Official Python and Django client SDK for FetchNode — the focused operations workspace for developers and teams managing multiple production Django websites.

FetchNode captures unhandled exceptions, error logs, slow database queries, N+1 query patterns, Celery background tasks, and authoritative business conversions with zero bloat and privacy-safe defaults.

---

Table of Contents

• Installation

• Quickstart (Django)

• Configuration Options

• Telemetry & Usage

  • Conversion & Lead Tracking
  • Automatic Error Monitoring
  • Manual Exception Capture
  • Celery Job Monitoring
  • Logging Handler

• Model Context Protocol (MCP) for AI Agents

• Privacy & Security Defaults

• Testing in Development

• Compatibility

• Links & Resources

• License

---

Installation

Version 0.5.1 adds error triage, performance bottlenecks, and SEO optimization to the Model Context Protocol (MCP) server. See the SDK changelog for changes and upgrade instructions.

Install via pip:

pip install fetchnode

Or using uv:

uv add fetchnode

(For Celery support, you can optionally install fetchnode[celery])

---

Quickstart (Django)

1. Add to INSTALLED_APPS

In your Django settings.py:

INSTALLED_APPS = [
    # ...
    "fetchnode",
]

2. Add the Middleware

Add FetchNodeMiddleware near the top of your MIDDLEWARE list for comprehensive request, query, and exception coverage:

MIDDLEWARE = [
    "fetchnode.middleware.FetchNodeMiddleware",
    "django.middleware.security.SecurityMiddleware",
    # ...
]

3. Configure settings.FETCHNODE

import os

FETCHNODE = {
    "API_KEY": os.environ.get("FETCHNODE_API_KEY", ""),
    "API_URL": "https://fetch-node.com/api/events/ingest/",
    "ENVIRONMENT": "production",  # e.g. "production", "staging"
}

Set FETCHNODE_API_KEY to your private server key in the deployment environment. Keep it out of source control and browser code. Restart or redeploy Django, then verify a request and test error in FetchNode.

Django will automatically initialize FetchNode upon startup and capture unhandled exceptions, slow queries, and request telemetry.

---

Configuration Options

Configure via settings.FETCHNODE or environment variables:

Setting Key	Environment Variable	Default	Description
API_KEY	FETCHNODE_API_KEY	""	Required. Your private server key from project settings. Never use a public browser key here.
ENDPOINT / API_URL	FETCHNODE_ENDPOINT	https://fetch-node.com/api/events/ingest/	Processing server URL. Point to your self-hosted instance if applicable.
ENVIRONMENT	FETCHNODE_ENVIRONMENT	"development"	Active environment name ("production", "staging").
RELEASE	FETCHNODE_RELEASE	""	Release tag or git commit hash (e.g. git rev-parse HEAD).
ENABLED	FETCHNODE_ENABLED	True if API_KEY present	Master toggle for event sending.
CAPTURE_TRANSACTIONS	—	True	Track request duration and database query counts.
CAPTURE_BODY	—	False	Privacy safeguard: request bodies are excluded by default.
CAPTURE_HEADERS	—	False	Privacy safeguard: request headers are excluded by default.
QUERY_SLOW_THRESHOLD_MS	—	100	Minimum execution time (ms) to flag an ORM query as slow.
N_PLUS_ONE_THRESHOLD	—	5	Identical SELECT query repetition threshold to flag N+1 issues.
SENSITIVE_KEYS	—	(built-in list)	Extra field names to scrub from payloads (additive).

---

Telemetry & Usage

Conversion & Lead Tracking

Track authoritative business events server-side (form submissions, payments, lead captures) with automatic session and visitor cookie attribution:

import fetchnode

def contact_view(request):
    form = ContactForm(request.POST)
    if form.is_valid():
        lead = form.save()

        # Track authoritative conversion
        fetchnode.track(
            "lead_submitted",
            request=request,
            idempotency_key=f"lead:{lead.id}",
            properties={"source": "contact_form"},
            conversion={"display_name": "Lead submitted", "funnel_order": 1, "primary": True},
        )
        return redirect("thank_you")

Why idempotency_key matters: Providing an idempotency_key (such as the database record ID or payment provider ID) ensures that retried requests, webhook replays, or browser reloads do not inflate your conversion metrics. FetchNode securely hashes the key client-side to produce a deterministic event ID.

Automatic Error Monitoring

Unhandled exceptions in Django views are caught by FetchNodeMiddleware, decorated with:

• The exact failing line of code and stack trace.
• In-app vs. framework/library frame classification.
• Database query count and execution time during the failing request.
• Correlated request ID and trace ID.

Manual Exception Capture

Capture handled exceptions or background exceptions:

import fetchnode

try:
    external_service_call()
except ThirdPartyAPIError as exc:
    fetchnode.capture_exception(exc, tags={"provider": "stripe"})

Celery Job Monitoring

FetchNode automatically instruments Celery when installed. You can also explicitly decorate tasks:

from fetchnode import capture_job

@capture_job(name="sync_external_inventory")
def sync_inventory():
    ...

Logging Handler

Send standard logging.ERROR and logging.CRITICAL records to FetchNode:

LOGGING = {
    "version": 1,
    "disable_existing_loggers": False,
    "handlers": {
        "fetchnode": {
            "level": "ERROR",
            "class": "fetchnode.handlers.FetchNodeHandler",
        },
    },
    "loggers": {
        "django": {
            "handlers": ["fetchnode"],
            "level": "ERROR",
            "propagate": True,
        },
    },
}

---

Model Context Protocol (MCP) for AI Agents

FetchNode includes a native Model Context Protocol (MCP) server. When pair programming or maintaining websites with AI agents (Claude Desktop, Cursor, Windsurf, Cline, or Python agents), your agents can query live Google Search Console rankings and technical error health directly:

python manage.py fetchnode_mcp --setup

This single command configures your editor agent and automatically documents the available tools in AGENTS.md / CLAUDE.md.

Available MCP Tools

• get_seo_overview: Organic clicks, search views, CTR, average position, and period changes.
• get_keyword_opportunities: High-potential striking-distance queries (positions 4–20), low-CTR terms, and zero-click keywords.
• get_page_performance: Landing page traffic, conversions, conversion rate, and driving queries.
• get_declining_pages: Landing pages and search queries losing organic Google clicks.
• get_technical_seo_health: Open 500 server errors and 404s affecting landing paths.

In-Code Python Usage

You can also query MCP tools directly in Python (e.g. within Celery tasks, LangChain, CrewAI, or autonomous scripts):

from fetchnode.mcp import get_seo_overview, get_keyword_opportunities

overview = get_seo_overview(period_days=28)
opportunities = get_keyword_opportunities(category="striking_distance")

---

Privacy & Security Defaults

FetchNode is engineered with a strict privacy-first architecture:

1. Client-side Scrubbing: Sensitive keys (password, token, secret, api_key, authorization, cookie, csrfmiddlewaretoken) and any compound keys like user_password or client_secret are scrubbed before payload transmission.
2. Safe SQL Normalization: Raw SQL text is stripped of literals and parameters (SELECT * FROM user WHERE id = 123 becomes SELECT * FROM user WHERE id = ?).
3. No Unintentional PII: Request bodies, query strings, and raw IP addresses are excluded by default.
4. Inspectable & Lightweight: The entire client is open-source Python with zero compiled binaries and zero unnecessary dependencies.

---

Testing in Development

In your local test suite (e.g. pytest or python manage.py test), FetchNode can be disabled by omitting the API key or explicitly setting:

# settings_test.py
FETCHNODE = {
    "ENABLED": False,
}

To test that event transmission works in your staging environment, run:

import fetchnode
fetchnode.capture_message("FetchNode verification ping", level="info")
fetchnode.flush()

---

Compatibility

• Python: 3.10, 3.11, 3.12, 3.13, 3.14
• Django: 4.2 LTS, 5.0, 5.1, 5.2
• Operating Systems: Linux, macOS, Windows

---

Links & Resources

• Website & Cloud App: fetchnode.com
• Documentation & Setup Guide: fetchnode.com/guide/
• Changelog: fetchnode.com/changelog/
• GitHub Repository: github.com/remcovanwijk040/fetchnode
• Issue Tracker: github.com/remcovanwijk040/fetchnode/issues

---

License

MIT License. See LICENSE for details.

SDK changelog

0.5.1 — 2026-10-16

Added

• Error management MCP tools: list_errors, get_error_details, resolve_error, and ignore_error (with feedback and snooze support).
• Performance bottleneck analysis MCP tool: get_performance_bottlenecks analyzing p95 latency, DB overhead %, N+1 query patterns, and actionable Django recommendations.
• Actionable SEO advice MCP tool: get_seo_optimization_advice providing prioritized actions for striking-distance keywords, CTR optimization, and declining pages.
• Standard MCP resources and prompts for error triage, performance bottlenecks, and SEO audits.
• Python SDK helpers in fetchnode.mcp.

Upgrade

python -m pip install --upgrade fetchnode==0.5.1

0.5.0 — 2026-10-16

Added

• Model Context Protocol (MCP) support for AI coding agents (Cursor, Windsurf, Claude Desktop, Cline).
• Django management command python manage.py fetchnode_mcp with --setup [all|cursor|windsurf|cline|claude], --print-config, --remote, and --status/--test.
• Automatic guideline documentation injection into repository agent prompt files (AGENTS.md, CLAUDE.md, .cursorrules).
• Direct Python SDK helper functions in fetchnode.mcp and fetchnode_client.mcp (get_seo_overview, get_keyword_opportunities, get_page_performance, get_declining_pages, get_technical_seo_health).
• Stdio-to-remote JSON-RPC runner (fetchnode.mcp.run_stdio_server).

Upgrade

python -m pip install --upgrade fetchnode==0.5.0

0.4.3 — 2026-09-29

Fixed

• Use https://fetch-node.com/api/events/ingest/ as the default ingest endpoint.
• Let Django automatically discover the SDK AppConfig when fetchnode is in INSTALLED_APPS, so startup installs logging and enabled instrumentation.
• Preserve automatic AppConfig discovery for the legacy fetchnode_client name.
• Correct the documentation URLs to the active FetchNode website.

Documentation

• Configure the private server key through FETCHNODE_API_KEY and supply an explicit ingest URL. Keep private keys out of browser code and source control.
• Show explicit server conversion declarations and non-personal event properties.
• Verify reception after restarting or redeploying Django.

Release checks

• Test the installed wheel outside the repository, including Django AppConfig and automatic logging setup.
• Exercise requests, exceptions, error logs and idempotent conversions against a local FetchNode collector, with project isolation checks.
• Check distribution metadata and the release tag before uploading to PyPI.

Upgrade

python -m pip install --upgrade fetchnode==0.4.3

Update your dependency file or lockfile, restart or redeploy Django, and verify a request and test error in FetchNode. Existing installations can set API_URL explicitly if they use their own collector.