Open Data

Free censorship data for researchers, journalists, developers, and civil society.

CC BY 4.0Updated DailyNo Auth RequiredMachine Readable

Global Censorship Index

Primary dataset

● Live

Country-level internet censorship scores derived from 25.3M+ multi-source measurements (OONI, IODA, CensoredPlanet). Includes block rates, ISP breakdowns, protocol distribution, and trend analysis.

Countries

127

Measurements

25.3M+

Update

Daily

Sources

Multi-source

JSON CSV HuggingFace Browse Methodology

Schema: country, score, level, trend, samples, blocked_count, block_rate, isps[]

Historical Archive

10-year measurement archive

● 2.2B+ source measurements

Complete 10-year archive of OONI measurements (2014-2024). 1.6M aggregated records covering 127+ countries. Ideal for longitudinal research and ML training.

Records

1.6M

Raw Measurements

2.2B+ (OONI aggregate)

Time Range

10 years

Format

Parquet

HuggingFace

Schema: date, country, anomaly_count, ok_count, confirmed_count, failure_count, measurement_count, anomaly_rate

ISP Breakdown

Granular ISP-level data

Per-ISP censorship analysis including ASN identifiers, block rates by provider, and specific blocked domains detected through multi-source measurements.

ISP Risk Index API Docs

Schema: country, asn, name, block_rate, threat_level, blocked_domains[]

7-Day Risk Forecast

Predictive censorship intelligence

New

ML-powered predictions of censorship risk for the next 7 days. Uses historical patterns, election calendars, and political events to forecast shutdowns.

Countries

127+

Horizon

7 days

Update

Hourly

Events

Wikipedia + GDELT

API Docs Election Risk Docs

Response: forecast[], summary (max_risk, drivers), confidence

VPN Accessibility

Unique data only Voidly provides

New

Real-time VPN reachability data from 37+ global probe nodes. Answers the question: "Can users in Iran/China/Russia actually connect to VPNs?"

Probe Nodes

37+

VPN Providers

Update

5 min

Unique

Yes ✓

API Docs

Tests: Voidly nodes, NordVPN, ProtonVPN, Mullvad (WireGuard UDP:51820)

Block Page Fingerprinting

Attribution of blocking entities

New

When we detect a block, we identify who is blocking. Distinguishes between government-mandated (Roskomnadzor, Iran TIC) vs ISP filters (Fortinet, Netsweeper).

Fingerprints

15+

Countries

10+

Entities: Roskomnadzor (Russia), Iran TIC, Great Firewall (China), BTK (Turkey), CITC (Saudi), PTA (Pakistan), Kominfo (Indonesia), Fortinet, Netsweeper, and more

Voidly Network Data

Real-time HTTP blocking from 37+ global nodes

New

Domain-level blocking data from the Voidly Network. 62 domains tested every 5 minutes from 37+ nodes (core + community + Fly.io). Includes blocking method (dns-poison, tcp-reset, blockpage) and blocking entity attribution (Roskomnadzor, Iran TIC, etc.).

Probe Nodes

37+

Domains

Interval

5 min

Attribution

Yes

Network Status Probe Stats Run a Probe API Docs

Run your own probe node:

pip install voidly-probePyPI|Docker Hub

Endpoints: /v1/probe/network, /v1/probe/stats, /v1/probe/domain/{domain}

Platform Risk Scores

Per-platform censorship intelligence

New

Censorship risk scores for major platforms (Twitter, WhatsApp, Telegram, YouTube, etc.) broken down by country. Answers: "How censored is platform X in country Y?"

Explore API Docs

Endpoints: /v1/platforms/scores, /v1/platforms/{platform}, /v1/platforms/country/{code}

Service Accessibility

Real-time accessibility oracle

New

Can users in country X access service Y right now? Real-time checks backed by37+ probe nodes, OONI, CensoredPlanet, and IODA. Designed for circumvention tools and researchers.

Try It API Docs

Endpoints: /v1/accessibility/check, /v1/accessibility/batch, /v1/accessibility/service/{name}

MCP Server — 83 Tools

For Claude, Cursor, Windsurf, Cline & more

● Live on npm

Model Context Protocol server with 83 tools covering censorship intelligence, Voidly Network monitoring, and E2E encrypted agent relay.

npx @voidly/mcp-server

npm GitHub Explore All 83 Tools →

27 censorship tools + 56 agent relay tools · v2.9.1 · CC BY 4.0

Python SDK

For LangChain, CrewAI & custom AI agents

New

E2E encrypted agent-to-agent communication. Async messaging, channels, memory, tasks, trust scoring. Built-in LangChain and CrewAI integrations.

pip install voidly-agents[all]

PyPI GitHub Agent Relay Docs →

40+ methods · LangChain (9 tools) · CrewAI (7 tools) · MIT License

API Access

GET /data/meta

Dataset versioning & endpoint discovery

GET /data/censorship-index.json

Full dataset with schema.org markup

GET /data/incidents

ML-classified censorship incidents (5,700+ events, 33,600+ evidence items, growing daily)

GET /data/incidents/delta?since=...

Delta feed - changes since timestamp (for sync)

GET /data/country/:code

Single country data (e.g., /data/country/CN)

GET /v1/vpn-accessibility

✨ VPN reachability from 37+ global nodes (unique data)

GET /data/methodology

Machine-readable methodology

POST /hydra/v1/predict

ML censorship prediction (99.8% F1, internal eval)

Full API Documentation →

Citation

Voidly Research. (2026). Global Censorship Index [Data set]. https://voidly.ai/live

License: CC BY 4.0 • Free to use with attribution

Data Sources

PrimaryOONI (Open Observatory of Network Interference)

OutagesIODA (Internet Outage Detection & Analysis)

DNS/HTTPCensoredPlanet (University of Michigan)

Probes37+ nodes (core + community + Fly.io, 5 min intervals)

ML Model99.8% F1 classifier (GradientBoosting v2, internal eval)

Update FrequencyDaily @ 03:00 UTC

Use Cases

Researchers

Academic studies on internet freedom

Journalists

Reporting on censorship events

Developers

Building circumvention tools

Reference Data

Structured censorship data for any downstream use