seojeom-mcp
Local-first MCP server that binds to a project opened by the Seojeom desktop app and serves its local wiki and graph data over stdio, enabling project-aware reading, searching, and writing of wiki and graph content.
README
Seojeom MCP Server
Canonical repo location: packages/seojeom-mcp
During the Godot-only decommission transition, apps/desktop/mcp-server may already be absent. New work should treat this package path as the source of truth. If a deprecated local lane still needs the old path to exist, recreate the minimal metadata-wrapper with pnpm apply:legacy-compat-wrapper.
Local-first MCP server for Claude and Codex.
seojeom-mcp binds to the project currently opened by the Seojeom desktop app or Godot launcher through a shared registry directory, then serves that project's local wiki and graph data over stdio.
Quickstart
Claude Code
claude mcp add -s local \
-e SEOJEOM_SHARED_REGISTRY_DIR=/mnt/c/Users/<you>/.seojeom/registry \
seojeom -- npm exec --yes --package=seojeom-mcp seojeom-mcp -- \
--router \
--host-kind auto \
--approval-mode prompt
Claude Desktop
Add this to claude_desktop_config.json and restart Claude Desktop:
{
"mcpServers": {
"seojeom": {
"command": "npm",
"env": {
"SEOJEOM_SHARED_REGISTRY_DIR": "<shared-registry-dir>"
},
"args": [
"exec",
"--yes",
"--package=seojeom-mcp",
"seojeom-mcp",
"--",
"--router",
"--host-kind", "auto",
"--approval-mode", "prompt"
]
}
}
}
Print the exact onboarding command first
npm exec --yes --package=seojeom-mcp seojeom-mcp -- --print-claude-onboarding
What it provides
- project binding to the currently active desktop/Godot project
- local wiki read/search/write tools
- local graph read/search/query tools
- core graph mutation/proposal tools for local authoring tasks
- approval-gated mutating operations
The exact tool surface is discovered at runtime through MCP tools/list.
Runtime contract
- primary transport:
stdio - primary install path:
npm exec --yes --package=seojeom-mcp seojeom-mcp -- - recommended mode: shared-registry router mode
- expected runtime: Node
>=20.20.0 - public npm package shape: standalone stdio entrypoint only
Source build contract
Running pnpm build in this package is a dual-output build:
dist/index.jsand the rest ofdist/**stay available as the local source/runtime HTTP sidecar surface used by Godot and other repo-local lanes.public-package/dist/standalone.jsis the staged npm publish surface
After a successful build, you can verify both lanes with:
pnpm smoke:build-layout
Godot-side source verification can also run the local HTTP sidecar smoke directly from the canonical package root:
pnpm smoke:sidecar-health
Surface tiers
seojeom-mcp is not one uniform surface. Treat these as separate contracts:
1. npm public package
- audience: external Claude/Codex users installing from npm
- transport:
stdio - mode: shared-registry router
- published shape:
LICENSE,README.md,bin/seojeom-mcp,dist/standalone.js,package.json - capability promise: public tools only
- non-goals: internal prompts, internal resources, full authoring harness, desktop UI shell
This is the only contract guaranteed by the published npm package.
2. local full runtime
- audience: local maintainers running from source
- transport: stdio plus local sidecar HTTP runtime paths
- profile: full authoring surface
- includes: prompts, resources, specialized graph/wiki authoring flows, internal playbooks
This source-tree runtime is intentionally larger than the public package.
3. desktop-coupled runtime
- audience: local desktop/Godot-integrated operation
- product default: app-bundled sidecar runtime
- depends on: shared registry writer, approval queue UI, review shell, sidecar health surface
- examples: approval review pane, desktop approval alerts, graph IPC bearer-token flows
- debug override: sidecar entry override is reserved for debug-only startup paths
These desktop-coupled capabilities are not part of the standalone npm guarantee.
Registry
- MCP Registry identifier target:
io.github.seojeom/seojeom-mcp - planned publish mode: GitHub Actions Trusted Publishing +
npm publish --provenance
Notes
- This package is still a preview release and the tool surface may evolve between versions.
- Router mode is the recommended public path. Direct per-project standalone flags remain as a compatibility fallback.
- The public package defaults to a core authoring tool surface. Prompts, resources, and more specialized orchestration surfaces are reserved for non-public/internal profiles.
- Public package documentation intentionally stays focused on installation and end-user usage. Internal release automation and repository-specific operator workflows are not part of the package contract.
Links
- Repository:
https://github.com/seojeom/seojeom-mcp - Issues:
https://github.com/seojeom/seojeom-mcp/issues - License: MIT
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.