MLT-OSS/FirstData

Free

全球最全面的权威数据源知识库，132+ 经验证数据源（政府、国际组织、学术机构），帮助 AI 减少幻觉。提供结构化元数据、100% URL 验证、中英双语支持。目标：1000+ 数据源。

FreeFree tier

Type

Open Source

Visit Website GitHub (40 stars)

About MLT-OSS/FirstData

FirstData is an open-source knowledge base that aggregates and structures authoritative data sources from governments, international organizations, and academic institutions worldwide, designed as an 'External Facts Context Layer' for AI Agents. It currently includes over 132 verified data sources with plans to reach 1000+. Key features include structured metadata with authority level classification, 100% URL verification, AI-powered intelligent search, MCP protocol integration for seamless AI agent access, Agent Skill distribution for zero-touch setup, and bilingual (Chinese/English) support. FirstData aims to provide a trusted foundation for AI models to reduce hallucinations, enable deep research with computable evidence chains, and upgrade AI reasoning from vague summarization to rigorous citation of original evidence.

Key Features

132+ verified data sources from government, international organizations, and academia

Structured metadata system with unique ID, description, access paths, API info, and authority level

Authority level classification (government, international, research, market, commercial, other)

AI-powered intelligent search using LLM to understand complex queries

MCP protocol integration for easy connection to Claude Desktop, Cline, and other AI apps

Agent Skill distribution for automatic registration and MCP configuration

Bilingual support (Chinese and English) for all metadata

100% URL verification ensuring every link is tested and available

Deep coverage of Chinese data sources – globally unique

Planned expansion to 1000+ data sources

Pros & Cons

Pros

Comprehensive coverage of authoritative data sources across multiple domains
Structured and machine-readable metadata for easy programmatic access
Open source and completely free to use
Built-in MCP support simplifies integration with AI applications
Bilingual metadata lowers language barriers for global users
100% URL verification ensures reliability and prevents broken links
AI-powered search makes it easy to find relevant data sources

Cons

Currently limited to 132+ data sources (goal of 1000+)
Requires technical knowledge to set up MCP server and integrate
Data source coverage may not include all niche domains yet
Documentation is primarily in Chinese with partial English

Best For

Provide anti-hallucination and anti-poisoning trusted foundation for AI modelsEnable deep research with computable and revisable evidence chainUpgrade AI from vague summarization to rigorous reasoning and citation based on original evidenceIntegrate authoritative external fact context into AI agents via MCP protocolAssist researchers in finding and accessing government and international data sources

FAQ

What is FirstData?

FirstData is an open source knowledge base that aggregates and structures authoritative data sources (government, international organizations, academic) into a machine-readable format, designed as an external facts context layer for AI agents to reduce hallucinations and provide verifiable evidence.

How does FirstData reduce AI hallucinations?

By providing a structured, verified, and authoritative set of data sources with 100% URL verification, authority level classification, and complete metadata, AI models can reference original evidence instead of relying on vague second-hand information.

What types of data sources are included?

FirstData includes government agencies, international organizations (World Bank, IMF, OECD, WHO), academic institutions (NBER, PubMed), and industry authorities across domains like economics, health, climate, energy, and more.

Is FirstData free and open source?

Yes, FirstData is completely free and open source under the MIT license, hosted on GitHub.