metagraphed

metagraphed

MCP server that enables AI agents to explore Bittensor subnets, check their health and economics, and discover their public APIs and schemas.

Category
访问服务器

README

<div align="center">

<a href="https://metagraph.sh"><img src="https://raw.githubusercontent.com/JSONbored/metagraphed/main/public/brand/banner-readme-mint.png" alt="Metagraphed — Bittensor subnet operational layer · data hub · API" width="820"></a>

Every subnet, metagraphed.

Bittensor didn't ship with a map, so we drew one. For every subnet: what does it expose (public APIs, docs, schemas), is it healthy, and how do I call it — machine-readable and agent-native from the ground up.

Website MCP MCP Registry mcp.so smithery badge npm PyPI License: AGPL-3.0 Gittensor impact

Website  ·  API  ·  OpenAPI  ·  GraphQL  ·  MCP  ·  Agent docs  ·  Agent workflows  ·  Feeds  ·  npm  ·  PyPI

<a href="https://gittensor.io/miners/repository?name=JSONbored/metagraphed"> <img src="https://raw.githubusercontent.com/JSONbored/metagraphed/gittensor-impact-assets/gittensor-impact-dark.svg" alt="Gittensor contributor impact" width="600"> </a>

<sub>Gittensor contributor impact — refreshed weekly by .github/workflows/gittensor-impact.yml</sub>

</div>

What it is

The native Bittensor metagraph tells you what's happening at the protocol layer. Metagraphed adds the builder-facing layer it lacks: a chain-direct block explorer (its own Rust indexer, no third-party RPC dependency), a registry of every subnet's public interfaces, endpoint health, and machine-readable schemas, and an agent toolkit — 200+ MCP tools, paste-ready OpenAI/Anthropic specs, a drop-in skill — built for integration developers, often reached through their AI agents. A live read-only RPC/WSS proxy and non-custodial, wallet-signed staking/delegation round it out.

Not an official OpenTensor/Bittensor project · not a replacement for the native metagraph · not an alpha/price dashboard · not custodial — staking and delegation are wallet-signed client-side; keys and credentials never touch our backend.

The web UI lives at metagraph.sh. The API is served from https://api.metagraph.sh (REST under /api/v1, artifacts under /metagraph).

Quickstart

Three ways to use Metagraphed. Pick one.

🤖 AI agent (MCP)

Agent-native, public, read-only, Streamable-HTTP. 200+ tools to explore the chain, discover a subnet, check if it's up, read its economics and metagraph, trace what a wallet does across the network, and learn how to call it.

claude mcp add --transport http metagraphed https://api.metagraph.sh/mcp

Cursor / other clients: add an MCP server with url https://api.metagraph.sh/mcp, transport streamable-http.

Tools: search_subnets · list_subnets · find_subnets_by_capability · get_subnet · get_subnet_health · get_subnet_economics · get_subnet_trajectory · get_subnet_metagraph · list_subnet_validators · get_neuron · get_account · get_account_events · get_account_subnets · list_subnet_apis · get_api_schema · get_fixture · get_agent_catalog · get_best_rpc_endpoint · registry_summary · list_enrichment_targets · find_subnet_opportunities · semantic_search · ask · find_subnet_for_task · how_do_i_call · verify_integration

📦 Typed client

Generated from the OpenAPI contract, published with provenance.

npm i @jsonbored/metagraphed   # JS/TS
pip install metagraphed        # Python

🌐 REST

Stable JSON envelope { ok, data, meta, error }. OpenAPI at /metagraph/openapi.json.

curl https://api.metagraph.sh/api/v1/subnets

🔮 GraphQL

Shape one request across the registry — a subnet with its health, surfaces, endpoints, and economics, a provider with its subnets, or the economic opportunity boards. POST to /api/v1/graphql; GET returns the SDL and introspection is enabled.

curl -X POST https://api.metagraph.sh/api/v1/graphql \
  -H 'content-type: application/json' \
  -d '{"query":"{ subnet(netuid: 7) { name health { status } surfaces { kind url } economics { emission_share } } }"}'

For agents

Resource URL
Copyable agent prompt /agent.md
Agent workflows /agent-workflows.md
Machine index /llms.txt
GraphQL /api/v1/graphql — POST a shaped query (subnet + health + surfaces + endpoints + economics, provider + subnets, opportunity boards); GET returns the SDL
Drop-in skill /skills/bittensor/SKILL.md — or gh skill install JSONbored/metagraphed bittensor
Resources index /metagraph/agent-resources.json
Content feeds /api/v1/feeds/registry — registry changes + incidents, as RSS / Atom / JSON Feed; /api/v1/feeds/gaps for ranked coverage gaps (per-subnet at /api/v1/feeds/subnets/{netuid})
Embeddable badge ![metagraphed](https://api.metagraph.sh/api/v1/subnets/{netuid}/badge.svg) — SVG (also /providers/{slug}/badge.svg); ?metric=uptime for reliability, ?metric=completeness for coverage score, plus ?style=flat-square and ?label=…

This repo

Cloudflare Worker API + Node build scripts. Schema-first: JSON Schema is the canonical contract → OpenAPI → types/clients. Artifacts are deterministic JSON; data refreshes on a schedule to R2/KV.

docs/              product + operating notes (start here)
registry/          subnet overlays, candidates, community submissions
schemas/           canonical JSON Schema components
scripts/           validation, generation, probe, safety
workers/           Cloudflare Worker API routes
public/metagraph/  compact generated artifacts + contracts
generated/         generated TypeScript types + client

Deeper docs: docs/api-stability.md (the /api/v1 contract), docs/curation-playbook.md.

Contributing

Open the repo in a devcontainer-aware tool (VS Code, GitHub Codespaces, the devcontainer CLI) for a zero-setup environment — Node 22 + Playwright's Chromium preinstalled. Issues are labeled good first issue and help wanted — start there.

  • Schema-first edits require npm run build (regenerates openapi.json + types).
  • Community submissions are PR-first: add a surface to exactly one registry/subnets/<slug>.json file (via npm run surface:add; use npm run subnet:new first when that subnet has no file yet), no generated artifacts.

See CONTRIBUTING.md and docs/curation-playbook.md. Gittensor contributor impact is pinned at the top of this README.

Subnet catalog

<!-- BEGIN:REGISTRY-CATALOG -->

128 curated subnets — 117 with a site, 46 with docs, 117 with a public repo. Live health, search, and the full list (every active subnet, not just the curated ones) at metagraph.sh; per-subnet JSON at https://api.metagraph.sh/api/v1/subnets/{netuid}. Root (SN0) is base-layer chain infrastructure, listed separately below, not counted as a subnet.

Focus areas: compute 9 · data 7 · defi 7 · inference 5 · language-models 4 · prediction-markets 4 · computer-vision 3 · finance 3 · dashboard 2 · data-artifact 2 · decentralized-training 2 · depin 2

<sub>Auto-generated from the curated overlays in registry/subnets/ by scripts/generate-registry-readme-section.ts — enrich a subnet (one PR) and it appears here. Not the live list; browse + monitor everything at metagraph.sh.</sub>

<!-- END:REGISTRY-CATALOG -->

Related

  • Frontend — apps/ui/ in this repo (folded in from the standalone metagraphed-ui repo via monorepo consolidation): the web app at metagraph.sh — block explorer, subnet directory, and a non-custodial staking/delegation flow (/delegate). Vite + React 19 + TanStack Start, deployed as its own independent Cloudflare Worker. Holds no subnet data — it renders what this backend serves.

License

The backend (Cloudflare Worker + build pipeline) is AGPL-3.0. The published client SDKs are permissively licensed so you can embed them freely — packages/client (npm) and python/ (PyPI) are Apache-2.0.

© 2026 JSONbored

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选