Cardano Unified MCP Server
An MCP server that provides AI assistants with searchable access to Cardano ecosystem documentation, SDKs, smart contract languages, governance, scaling, and developer standards from a single endpoint.
README
Cardano Unified MCP Server
An independent, community-maintained Model Context Protocol (MCP) server that gives AI assistants deep knowledge of the Cardano ecosystem — documentation, SDKs, smart contract languages, governance, scaling, and developer standards, all searchable from a single endpoint.
Run by Easy1Staking. Hosted instance: mcp.easy1staking.com
New here? MCP is an open standard that lets an AI assistant plug into external knowledge servers. Connect this one and your assistant can answer Cardano development questions with citations to real upstream docs instead of making things up.
Read ABOUT.md for the full story — what this project is, where the knowledge comes from, how it is vetted, and who maintains it. Read docs/architecture.md for the component, query, and ingestion diagrams.
What's Inside
50+ documentation sources across 8 categories, sourced from the cardano-dev-skills curated registry:
- Infrastructure — Ogmios, Kupo, Blockfrost, Mithril, Oura, Pallas, Dolos, Yaci Store, Koios, Cardano GraphQL, Cardano Wallet, Cardano Node Wiki, DB-Sync, Cardano CLI, Amaru
- Smart Contracts — Aiken (lang + stdlib + examples + design patterns), Plutus, OpShin, Plutarch, Plu-ts, Scalus, Pebble, Helios, CIP-113 Programmable Tokens (core + platform), Smart Contract Vulnerabilities, Cardano Use Case Templates
- SDKs — Mesh SDK, Evolution SDK, cardano-js-sdk, PyCardano, cardano-client-lib, Cardano Serialization Lib, Buildooor, Cardano Connect with Wallet, Cardano Addresses, CIP-113 SDK
- Standards — CIPs (170+), Developer Portal, Cardano Docs, Cardano Ledger
- Governance — GovTool, SanchoNet
- Scaling — Hydra, Ouroboros Leios
- Oracles — Charli3 (Pull Oracle Contracts/Client/SDK)
- Testing — Yaci DevKit, Cardano Node Antithesis
The authoritative source list lives in cardano-dev-skills/registry/sources.yaml. To add or remove a source, open a PR there.
Features
MCP Tools
search_docs— Hybrid semantic + keyword search across all indexed documentationget_doc— Retrieve full content of a specific documentlist_topics— Browse available sources and their topicslist_skills— List the workflow skills exposed by this serverget_skill— Retrieve the full SKILL.md for a named workflow
MCP Resources
cardano://sources— Overview of all indexed sourcescardano://source/{name}— Topic listing for a specific sourcecardano://doc/{source}/{path}— Full document content
MCP Prompts
All 15 cardano-dev-skills workflows are exposed as MCP prompts. Each takes an optional request argument with the user's specific context, then dispatches the skill's workflow (which uses search_docs for citations).
build-transaction · cardano-context · connect-wallet · debug-transaction · design-token · explain-cip · explain-eutxo · governance-guide · optimize-validator · query-chain · review-contract · scaffold-project · setup-devnet · suggest-tooling · write-validator
Quick Start
Use the hosted instance
Add to your MCP client configuration (Claude Desktop, Claude Code, Cursor, etc.):
{
"mcpServers": {
"cardano": {
"url": "https://mcp.easy1staking.com/mcp"
}
}
}
Run locally (stdio)
# Clone both repos as siblings — this server depends on cardano-dev-skills.
git clone https://github.com/easy1staking-com/cardano-dev-skills.git
git clone https://github.com/easy1staking-com/cardano-unified-mcp-server.git
cd cardano-unified-mcp-server
npm install
npm run build
# Ingest documentation (requires EMBEDDINGS_API_KEY for semantic search)
cp .env.example .env
# Edit .env with your OpenAI API key
npm run ingest
# Run in stdio mode
node dist/index.js --stdio
Add to your MCP client:
{
"mcpServers": {
"cardano": {
"command": "node",
"args": ["/path/to/cardano-unified-mcp-server/dist/index.js", "--stdio"]
}
}
}
Run as HTTP server
# Start the server (default port 3000)
npm start
# Or with API key protection
MCP_API_KEY=your-secret npm start
Configuration
| Variable | Default | Description |
|---|---|---|
PORT |
3000 |
HTTP server port |
HOST |
0.0.0.0 |
Bind address |
SKILLS_PATH |
../cardano-dev-skills |
Path to the cardano-dev-skills checkout (registry + vendored docs + skill workflows) |
EMBEDDINGS_API_KEY |
— | OpenAI API key for semantic search |
EMBEDDINGS_API_BASE |
https://api.openai.com/v1 |
OpenAI-compatible endpoint |
EMBEDDINGS_MODEL |
text-embedding-3-large |
Embedding model |
MCP_API_KEY |
— | Optional Bearer token for HTTP mode |
DB_PATH |
./data/docs.db |
SQLite database path |
Architecture
graph TB
subgraph clients["MCP clients"]
CD["Claude Desktop · Claude Code · Cursor · Windsurf · …"]
end
subgraph skills["cardano-dev-skills (sibling repo)"]
REG["registry/sources.yaml"]
VEND["docs/sources/<slug>/<br/>vendored markdown"]
SK["skills/*/SKILL.md<br/>workflow guides"]
end
subgraph server["Cardano Unified MCP Server"]
T["Tools (search_docs, get_skill, …)"]
R["Resources"]
P["Prompts (auto-loaded from skills)"]
DB[("VectorDB<br/>SQLite + FTS5 + sqlite-vec")]
T --> DB
R --> DB
end
REG --> server
VEND --> DB
SK --> P
SK --> T
clients -->|"stdio or HTTP + SSE"| server
Full component, query, and ingestion diagrams in docs/architecture.md.
Search modes
- Hybrid (default) — BM25 full-text (40%) + vector similarity (60%)
- Semantic — Embedding cosine similarity only
- Keyword — Full-text search with Porter stemming (no API key needed)
Kubernetes Deployment
The server is designed for stateless horizontal scaling on Kubernetes:
# Apply manifests
kubectl apply -f k8s/deployment.yaml
kubectl apply -f k8s/cronjob-ingest.yaml
# Create secrets
kubectl create secret generic cardano-mcp-secrets \
--from-literal=EMBEDDINGS_API_KEY=sk-... \
--from-literal=MCP_API_KEY=your-secret
- 2 replicas with health checks
- Persistent volume for the SQLite database
- Weekly CronJob for documentation re-ingestion (clones cardano-dev-skills HEAD first, then runs ingest)
Ingestion
# Ingest all sources from skills' registry
npm run ingest
# Ingest a specific source (name substring match)
npm run ingest -- Aiken
# Dry-run: read and validate files, but skip chunking and embeddings
npm run ingest -- --validate-only
# Skip embeddings (keyword-only mode, no API key needed)
npm run ingest -- --skip-embeddings
# Check that every registry entry resolves to a vendored dir
npm run validate:skills
The Skills Drift GitHub Action runs the contract check nightly against cardano-dev-skills HEAD — a red badge means the upstream registry has drifted in a way that would break Sunday's indexer.
Development
npm run dev # Watch mode with hot reload
npm run typecheck # Type checking only
npm run build # Compile TypeScript
License
Contributing
See CONTRIBUTING.md for guidelines and CODE_OF_CONDUCT.md for community standards. Note: new doc sources are added in cardano-dev-skills, not here.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。