| name | lumenloop-mcp-connect |
| description | Connect your AI assistant (Claude, ChatGPT, Gemini, Cursor, or any MCP client) to LumenLoop's free read-only Stellar ecosystem MCP at https://mcp.lumenloop.com and learn its 18 query tools. Use when the user wants to set up or troubleshoot the lumenloop connector, asks what Stellar ecosystem data is available, or needs to know which lumenloop skill fits a task. |
| user-invocable | true |
Connect to the LumenLoop Stellar Ecosystem MCP
This is the onboarding anchor for the LumenLoop skill family. It gets your AI
assistant connected to LumenLoop's free, read-only Stellar ecosystem MCP
(https://mcp.lumenloop.com), explains what data is behind it, lists the 18 query
tools, and routes you to the right sibling skill for the job. The server speaks the
open Model Context Protocol, so any MCP-capable client can connect — Claude Code,
Claude.ai, ChatGPT, Gemini, Cursor, or a custom agent. It is a public, read-only
window on the Stellar ecosystem: the project directory, Stellar Community Fund
(SCF) submissions, and editorial / news / AV / event / research content. Nothing
here writes data or costs anything — every tool just reads public information.
When to use this skill
- The user wants to set up the
lumenloop connector in Claude Code, Claude.ai, or
another MCP client.
- The connector is not showing up, tools fail, or sign-in is stuck.
- The user asks what Stellar ecosystem data is available through LumenLoop.
- You need a map of the 18 tools and their exact signatures.
- You are unsure which LumenLoop skill fits a task (scout vs dossier vs digest …).
- A teammate is starting fresh and needs the one-page orientation.
Related skills
- Map a sector into a landscape →
../stellar-ecosystem-scout/SKILL.md
- Deep-dive one project →
../stellar-project-dossier/SKILL.md
- Position an SCF idea →
../scf-submission-radar/SKILL.md
- Pick an ecosystem dependency →
../stellar-integration-finder/SKILL.md
- Dated, cited activity digest →
../stellar-ecosystem-digest/SKILL.md
- Turn an idea into a build path →
../stellar-builder-quickstart/SKILL.md
- Audit a draft against the ecosystem →
../stellar-content-auditor/SKILL.md
For building on Stellar (Soroban contracts, the JS stellar-sdk / Freighter /
Stellar Wallets Kit for dApps, classic assets + trustlines + SAC for tokens,
Stellar RPC / Horizon for chain data, SEP/CAP standards, the Reflector oracle), see the
official docs at https://developers.stellar.org.
Connect
Claude Code
This repo ships an .mcp.json that already declares the server, so if you are working
inside it the connector is preconfigured — just approve it when prompted. To add it
to any other project, run:
claude mcp add --transport http lumenloop https://mcp.lumenloop.com
Then complete the one-time sign-in:
/mcp # opens the MCP panel
# → select "lumenloop" → follow the prompt to sign in in your browser
After sign-in, the 18 tools appear as lumenloop tools. Verify with /mcp (status
should read connected) or by calling a no-arg tool such as get_categories.
Claude.ai, ChatGPT, Gemini & other MCP clients
- Open your client's connector settings (e.g. Claude.ai: Settings → Connectors →
Add; other clients have an equivalent "add MCP server / connector" screen).
- Add a custom / remote MCP connector with URL
https://mcp.lumenloop.com
(transport: streamable HTTP).
- Complete the one-time sign-in when prompted.
- The
lumenloop tools are now available in chat.
If your client cannot auto-load skill files, you can still use each
skills/<name>/SKILL.md as a pasted playbook/prompt — the recipes are plain
model-agnostic MCP tool-call sequences.
Smoke test
Call one tool with no arguments to confirm the connection is live:
get_categories()
→ {count: 5, categories: ["Applications", "Developer Tooling",
"Education & Community", "Financial Protocols", "Infrastructure & Services"]}
If that returns the five categories, you are connected.
The 18 tools at a glance
One line each below. Full per-tool signatures and parameters →
reference/tool-catalog.md.
Directory & taxonomy
| Tool | What it does |
|---|
search_directory | Keyword search projects by name/description/other_names (ILIKE). |
get_project | One project; compact: true for a slim writer shape. |
get_related_projects | Projects mentioned by a content item (content → projects). |
get_categories | Controlled category vocabulary (5 live values). |
get_regions | Distinct region values (based_in + operating_region). |
get_tags_vocabulary | Content tags (articles/AV) — not project tags. |
get_project_tags_vocabulary | Controlled project tags ("Wallet Infra", "SDK" …). |
Stellar Community Fund (SCF)
| Tool | What it does |
|---|
get_scf_submissions | SCF submissions for a project, by slug or fuzzy name. |
find_similar_scf_submissions | SCF submissions by embedding similarity (query xor slug). |
Semantic content (returns AI summaries + source url)
| Tool | What it does |
|---|
search_content_semantic | Semantic search across published content, by types. |
find_content_about_project | Content that discusses a given project. |
find_av_passages | Quotable passages inside AV transcripts. |
find_content_by_entity | Content mentioning a named entity (person/org/token/project). |
find_similar_projects_semantic | Projects semantically similar to a given project. |
Documents & research library
| Tool | What it does |
|---|
list_documents | Browse a collection (articles/events/videos/jobs/research). |
search_documents | Title/url search (alias of list_documents + search). |
get_document | Full details of one document. |
list_research | LumenLoop's published research library. |
Worked example — orient in 3 calls
Goal: confirm the connection and get the lay of the land before deeper work.
1. get_categories()
→ 5 buckets: Applications, Developer Tooling, Education & Community,
Financial Protocols, Infrastructure & Services
2. search_directory(query="wallet", limit=5)
→ {count: 5, projects: [{slug:"abroad", title:"Abroad", category:"Applications",
tags:[...]}, {slug:"adamik", ...}, {slug:"airgap", ...}, ...]} # short term → hits
3. find_similar_projects_semantic(slug="abroad", limit=5)
→ semantically adjacent wallet/payments projects to grow the cluster
From here, hand off: use stellar-ecosystem-scout to map the whole sector,
stellar-project-dossier to vet one of those slugs, or
stellar-integration-finder to choose one as a dependency.
Which skill for which job
| You want to… | Use skill | Anchor tools |
|---|
| Survey/map a sector or topic | ../stellar-ecosystem-scout/SKILL.md | search_directory, find_similar_projects_semantic, search_content_semantic |
| Vet / profile ONE project | ../stellar-project-dossier/SKILL.md | get_project, get_scf_submissions, find_content_about_project |
| Position an SCF idea vs prior art | ../scf-submission-radar/SKILL.md | find_similar_scf_submissions, get_scf_submissions |
| Pick a wallet/oracle/anchor/etc. dependency | ../stellar-integration-finder/SKILL.md | search_directory, get_project, find_similar_projects_semantic |
| Dated, cited recap of recent activity | ../stellar-ecosystem-digest/SKILL.md | list_documents, search_content_semantic, find_content_by_entity |
| Turn an idea into a build path | ../stellar-builder-quickstart/SKILL.md | search_directory, find_similar_projects_semantic + https://developers.stellar.org |
| Fact-check / enrich a draft (handles, citations) | ../stellar-content-auditor/SKILL.md | search_directory, get_project, find_content_by_entity |
Cross-cutting gotchas
search_directory is keyword/ILIKE, not semantic. SHORT terms ("wallet",
"oracle", "RWA") work; long natural-language phrases ("real world asset tokenization
platform") return 0. For concept/topic discovery use the semantic tools
(search_content_semantic, find_similar_projects_semantic) instead.
- Content tools return summaries, not full text.
search_content_semantic,
find_content_about_project, find_av_passages, and find_content_by_entity return
LumenLoop's AI-generated summaries plus the source url. Cite the url; do not
present a summary as the verbatim article.
search_content_semantic defaults to 5 hits per content type and to a
concise response (no long_summary). Raise limit or pass
response_format="detailed" when you need more depth; weak matches may carry a
_weak_match flag plus a _hint.
get_related_projects is content → projects. It takes content_id +
content_type and maps a piece of content to the projects it mentions — it is NOT
project → project. For project → project similarity use find_similar_projects_semantic.
get_project is heavy by default. It returns a large full row; pass
compact: true when you only need title / category / tags / links (~500 chars).
find_similar_scf_submissions wants query XOR slug. Pass a natural-language
query for topic search, or an existing submission slug (e.g. recsAQHCghw9HIUvF)
to find neighbors of that submission — not both.
- Vocabularies are controlled. Read the buckets first with
get_categories,
get_regions, get_tags_vocabulary, get_project_tags_vocabulary (these take no
args) so filters and tags match what the data actually uses.
- Tool missing after a server update? Reconnect the connector: in Claude Code
/mcp → lumenloop → re-authorize; in other clients, re-add / re-auth the connector.
Pointers