We use necessary cookies for site function. For analytics and marketing, we need your consent. You can change your mind anytime. Privacy Policy

Official docs, live ecosystem data, and proven playbooks — one connection, zero API keys
Stellar · AI · MCP

Stellar Raven: Give Your AI Agent the Official Docs, Live Ecosystem Data & Playbooks

Olvis GilOlvis Gil  ·  November 16, 2026  ·  6 min read

Stellar Raven is a remote MCP server that hands your AI coding assistant the whole Stellar knowledge base in one connection: official developer docs ranked for agents, live graded ecosystem data, community intel, and 20 tested playbooks it can pull mid-task. Paste one endpoint, sign in once in the browser, and the agent answers with sources — not vibes. If you already use Claude Code, Codex, Cursor, or VS Code for Stellar work, this is the missing piece.

What’s behind the endpoint

Four catalogs, cross-referenced: official docs (smart contracts, SDKs, RPC & Horizon, SEPs and standards, searched the way agents actually ask), live ecosystem data (920+ projects, 2,300+ graded repos, builder and partner intelligence refreshed from live services — not training-data stale), community intel (news, research, podcasts, events, jobs, governance and SCF funding flows), and 20 proven playbooks — step-by-step procedures for smart contracts, payments, dApps, and data/indexing that the agent reads section by section exactly when it needs them.

Under the hood: just two tools

Instead of dozens of tool schemas crowding your agent’s context window, Raven exposes two. search ranks its catalog of 62 live operations and 20 skills. execute runs agent-written JavaScript in a no-network sandbox where every call is validated against the catalog — sandboxed by design, so the agent can compose operations without your secrets ever reaching it. The catalog itself is checked against live services daily, which quietly solves the “my MCP server rotted when the API changed” problem.

Connect in two minutes

# Claude Code
claude mcp add --transport http stellar-raven \
  "https://raven.stellar.org/mcp"
# then /mcp -> Authenticate -> sign in in your browser

# Codex
codex mcp add stellar-raven --url "https://raven.stellar.org/mcp"
codex mcp login stellar-raven

# Cursor — ~/.cursor/mcp.json
{ "mcpServers": { "stellar-raven": { "url": "https://raven.stellar.org/mcp" } } }

# VS Code
code --add-mcp '{"name":"stellar-raven","type":"http","url":"https://raven.stellar.org/mcp"}'

# No native remote MCP? Bridge with mcp-remote
{ "mcpServers": { "stellar-raven": {
    "command": "npx",
    "args": ["-y","mcp-remote@latest","https://raven.stellar.org/mcp","--transport","http-only"] } } }

Why it beats the DIY pile

The alternative is a separate MCP server per service, API keys to mint and rotate, and you doing the cross-referencing when a question spans docs + ecosystem + standards. Raven collapses that to one endpoint and one sign-in, with answers composed across sources. There is also a web playground if you want to try the catalog before wiring it into your agent, and the code is on GitHub.

Pair it with the hands-on layer: Stellar Lab for click-through transaction building, the Stellar Go CLI — which itself ships an MCP server exposing wallet and payment tools — and the Stellar masterclass for structured learning.

Keep learning

Raven · Raven docs · Playground · Stellar hub · Bell, Protocol 30 · Devin AI masterclass

Frequently asked questions

What is Stellar Raven?
A remote MCP (Model Context Protocol) server run by Stellar at raven.stellar.org/mcp. It gives AI coding assistants — Claude Code, Codex, Cursor, Claude Desktop, VS Code — access to official Stellar docs, live graded ecosystem data, community intel, and 20 step-by-step playbooks through a single authenticated endpoint.
How do I connect my AI tool to Raven?
Add https://raven.stellar.org/mcp as a remote/streamable-HTTP MCP server in your tool — e.g. `claude mcp add --transport http stellar-raven https://raven.stellar.org/mcp` — then complete the browser OAuth sign-in. No service API keys are needed; tools without native remote-MCP support can bridge via mcp-remote.
What can my agent actually do with it?
Raven exposes two MCP tools: `search` ranks its catalog of 62 live operations and 20 skills/playbooks, and `execute` runs agent-written JavaScript in a sandboxed, no-network environment with calls validated against the catalog. Your agent gets cross-referenced answers — docs plus live ecosystem data plus playbooks — not single-source lookups.
Does it cost anything or need API keys?
One browser sign-in; no service API keys. Secrets stay server-side and the catalog is checked against live services daily, so answers reflect current network state rather than stale training data.
How is Raven different from just pasting docs into a prompt?
Coverage and freshness. A prompt gets fragments you paste; Raven gives the agent ranked access to the whole docs corpus, 920+ projects and 2,300+ graded repos of ecosystem intelligence, news/funding/governance intel, and executable playbooks — kept current daily.
Share this article
Olvis Gil
Olvis Gil

Founder of EduNode and OG Technologies EU — a Vienna consultancy for blockchain payments, tokenization, digital identity, and ISO 20022.