mcp-embedding-daemon 1.18.0b13

Shared ONNX/GGUF embedding server for the n24q02m MCP ecosystem

mcp-core

Shared foundation for building MCP servers -- Streamable HTTP transport, OAuth 2.1, browser-based credential setup, and a shared embedding daemon.

Sister projects from n24q02m (click to expand)

Project	Tagline	Tag
better-code-review-graph	Knowledge graph for token-efficient code reviews -- semantic search and call-...	MCP
better-email-mcp	IMAP/SMTP email for AI agents -- read, send, organize folders, and manage att...	MCP
better-godot-mcp	Composite MCP server for Godot Engine -- 17 composite tools for AI-assisted g...	MCP
better-notion-mcp	Markdown-first Notion for AI agents -- pages, databases, blocks, and comments...	MCP
better-telegram-mcp	Telegram for AI agents -- messages, chats, media, and contacts across both bo...	MCP
claude-plugins	Claude Code plugin marketplace for the n24q02m MCP servers -- install web sea...	Marketplace
imagine-mcp	Image and video understanding + generation for AI agents -- across Gemini, Op...	MCP
jules-task-archiver	Chrome Extension for bulk operations on Jules tasks via batchexecute API -- a...	Tooling
mcp-core	Shared foundation for building MCP servers -- Streamable HTTP transport, OAut...	MCP
mnemo-mcp	Persistent AI memory with hybrid search and embedded sync. Open, free, unlimi...	MCP
qwen3-embed	Lightweight Qwen3 text embedding and reranking via ONNX Runtime and GGUF	Library
skret	Secrets without the server.	CLI
tacet	TACET: a self-distilling neuro-symbolic cascade that amortises LLM cost in kn...	Tooling
web-core	Shared web infrastructure package for search, scraping, HTTP security, and st...	Library
wet-mcp	Open-source MCP server for AI agents: web search, content extraction, and lib...	MCP

Table of contents

• Packages
• What you get
• Quick start (Python)
• Quick start (TypeScript)
• Documentation
• Development
• License

mcp-core is the shared foundation for the n24q02m MCP servers: a Streamable HTTP transport, an OAuth 2.1 Authorization Server, lifecycle management, install automation, and a shared embedding daemon.

mcp-core is the functional successor to the archived mcp-relay-core. All crypto, storage, OAuth, relay, and schema modules from mcp-relay-core ship under the same paths in mcp-core (1:1 superset), so downstream MCP servers can migrate with a pure import + dependency rename. See the Migration guide for the rename table.

Packages

Package	Language	Registry	Install
packages/core-py	Python 3.13	PyPI: n24q02m-mcp-core	pip install n24q02m-mcp-core
packages/core-ts	TypeScript / Node 24	npm: @n24q02m/mcp-core	bun add @n24q02m/mcp-core
packages/embedding-daemon	Python 3.13	PyPI: mcp-embedding-daemon	pip install mcp-embedding-daemon

All three packages share the same version (semantic-release.toml bumps both Python pyproject.toml files plus the npm package.json in lockstep).

What you get

n24q02m-mcp-core (Python) and @n24q02m/mcp-core (TypeScript)

Identical public API in both languages:

• crypto/ — ECDH P-256, AES-256-GCM, HKDF-SHA256 primitives. Cross-language test vectors guarantee Python and TypeScript produce the same ciphertext for the same input.
• storage/ — PerPluginStore, the per-plugin encrypted credential store. Single-user (stdio / HTTP) writes ~/.<plugin>-mcp/config.json encrypted with a machine-bound key; HTTP multi-user writes ~/.<plugin>-mcp/subs/<sub>/config.json encrypted with a key derived from the CREDENTIAL_SECRET env var (salt <plugin>:<sub>). Pluggable CredentialBackends (LocalFsBackend, CfKvBackend) decouple the on-disk layout from serverless deployments. Also ships session lock files and config resolver helpers. The legacy shared config.enc file (storage.config_file) is deprecated.
• auth/ — the self-hosted OAuth 2.1 Authorization Server that downstream servers actually run: create_local_oauth_app (Starlette ASGI app serving /authorize, /token, the /.well-known/oauth-* metadata, and the browser-rendered credential form), render_credential_form, the optional shared-password gate at /login (MCP_RELAY_PASSWORD, empty disables it), and create_delegated_oauth_app for upstream-redirect / device-code multi-user flows.
• oauth/ — OAuth 2.1 primitives consumed by auth/: JWTIssuer (RS256), SqliteUserStore for multi-user mode, and OAuthProvider, the legacy mcp-relay-core PKCE-over-relay provider retained for migration.
• relay/ — RelaySession, create_session, poll_for_result, send_message plus the EFF Diceware wordlist for passphrase generation. This is the legacy mcp-relay-core ECDH relay-client path used by OAuthProvider; the live setup UX is the auth/ browser credential form.
• schema/ — RelayConfigSchema TypedDict that downstream servers use to declare their config form.
• transport/ — StreamableHTTPServer wrapper around FastMCP / @modelcontextprotocol/sdk Streamable HTTP transport, plus OAuthMiddleware (RFC 6750 + RFC 9728 compliant Bearer validation).
• lifecycle/ — LifecycleLock cross-platform file lock that prevents two server instances from binding the same (name, port) pair.
• install/ (Python only) — AgentInstaller that writes MCP server entries into Claude Code, Cursor, Codex, Windsurf, and OpenCode config files.

mcp-embedding-daemon

FastAPI HTTP server scaffold for the upcoming shared ONNX/GGUF embedding backend. Currently exposes:

• GET /health — returns {status, version}
• POST /embed — returns 501 with a roadmap link (backend wiring lands in the next release)
• POST /rerank — returns 501 with a roadmap link

CLI entry point: mcp-embedding-daemon --host 127.0.0.1 --port 9800.

Quick start (Python)

from mcp_core import RelaySession, create_session, decrypt
from mcp_core.transport.streamable_http import StreamableHTTPServer
from mcp_core.oauth import JWTIssuer
from mcp_core.transport.oauth_middleware import OAuthMiddleware
from fastmcp import FastMCP

mcp = FastMCP("my-server")

issuer = JWTIssuer("my-server")
issuer  # Use issuer.issue_access_token(sub) / verify_access_token(token)

middleware = [OAuthMiddleware(issuer=issuer, resource_metadata_url="http://127.0.0.1:9876/.well-known/oauth-protected-resource")]
server = StreamableHTTPServer(mcp, port=9876, middleware=middleware)
server.run()

Quick start (TypeScript)

import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js'
import { JWTIssuer } from '@n24q02m/mcp-core/oauth'
import { OAuthMiddleware, StreamableHTTPServer } from '@n24q02m/mcp-core/transport'

const server = new McpServer({ name: 'my-server', version: '0.0.0' })
const issuer = new JWTIssuer('my-server')
await issuer.init()

const middleware = new OAuthMiddleware({
  jwtIssuer: issuer,
  resourceMetadataUrl: 'http://127.0.0.1:9876/.well-known/oauth-protected-resource'
})

const http = new StreamableHTTPServer({ server, port: 9876, oauthMiddleware: middleware })
await http.connect()
// Then mount http.handleRequest(req, res) on your http.Server / Express / Hono.

Documentation

Full docs at mcp.n24q02m.com/servers/mcp-core/architecture/ (Foundation library section in the MCP n24q02m unified docs site):

• Architecture -- transport, OAuth AS, lifecycle, multi-user primitives
• Trust model -- threat model + key-handling guarantees
• Migration -- breaking-change history and upgrade paths from mcp-relay-core
• Shared services -- embedding daemon + ancillary docker-compose stack

Source of truth lives in n24q02m/claude-plugins/plugins/mcp-core/. Edit there; this repo's docs/ directory is intentionally minimal post-migration.

Development

mise run setup            # install runtimes + deps + pre-commit hooks
bun install               # root TypeScript workspace install

# Python (per package)
cd packages/core-py
uv sync --group dev
uv run pytest
uv run ty check
uv run ruff check .

# TypeScript
cd packages/core-ts
bun run test
bun run check
bun run build

License

MIT