Open MCP Knowledgebase
A zero-database MCP knowledgebase template for Vercel that exposes static Markdown content to AI agents through clean MCP surfaces.
README
Open MCP Knowledgebase
A zero-database MCP knowledgebase template for Vercel.
Use it to expose static Markdown content to AI agents through clean MCP surfaces. A surface is just a named content namespace with its own MCP endpoint, metadata, tools, search scope, resources, and optional provider integrations.
- Documentation, research, team knowledge, portfolio memory, client context, runbooks, or any other structured Markdown corpus.
- Multiple surfaces from one deploy, for example
docs,research,projects, or the shipped example surfacestechnologyandprojects. - Skill catalogs with
when_to_useguidance. - Build-time MiniSearch retrieval.
- Optional provider-backed integrations behind content metadata.
- Optional passthrough tools, such as
write_content, for deployment-level capabilities that proxy a configured external service.
The default deployment needs no database and no paid provider. Content is committed as Markdown, indexed at build time, and bundled with the Next.js MCP routes.
Quick Start
pnpm install
pnpm validate
pnpm build
pnpm dev
Set a local API key:
cp .env.example .env.local
Then connect an MCP client to:
http://localhost:3000/api/mcp/technologyhttp://localhost:3000/api/mcp/projects
Those are example surfaces from the starter content. Add or rename top-level folders under content/ to create your own surface names.
Use Authorization: Bearer <MCP_API_KEY> where the client supports headers.
Clients without header support can use ?api_key=<MCP_API_KEY> only when MCP_ALLOW_QUERY_API_KEY=true. Keep that fallback disabled for production unless you explicitly need it for a known connector.
Claude.ai custom connectors without OAuth are one known case: add the connector with /api/mcp/<surface>?api_key=<MCP_API_KEY> after enabling MCP_ALLOW_QUERY_API_KEY=true. See MCP client examples.
MCP Metadata
The server returns MCP implementation metadata during initialization:
MCP_SERVER_SLUGcontrols the stable machine-readableserverInfo.name.MCP_SERVER_NAMEcontrols the human-readable base title.MCP_PUBLIC_URL/MCP_WEBSITE_URLpopulateserverInfo.websiteUrl.MCP_ICON_URL,MCP_ICON_MIME_TYPE,MCP_ICON_SIZES, andMCP_ICON_THEMEconfigure client icons where supported.
Deployment Modes
Enable one or more surfaces with:
MCP_ENABLED_SURFACES=technology,projects
The valid values are discovered from top-level content/<surface>/_meta.yaml files. Single-surface downstreams are useful when you want stricter separation:
MCP_ENABLED_SURFACES=technology
MCP_ENABLED_SURFACES=projects
If MCP_ENABLED_SURFACES is unset, every discovered surface is enabled. If it is set, only recognized values are enabled; invalid or empty configured values expose no surfaces.
Commands
pnpm dev # Next.js dev server
pnpm build # Build indices, then build Next.js
pnpm test # Node test suite
pnpm typecheck # TypeScript
pnpm lint # Biome
pnpm validate # Content schema validation
pnpm stdio # Local stdio MCP, defaults to the first enabled surface
pnpm inspect:stdio # Launch MCP Inspector against the stdio server
pnpm sync:list # List sync sources
pnpm sync:free # Run free sync sources
Protocol Safety
The stdio entrypoint keeps stdout protocol-clean. Runtime diagnostics go to stderr so local MCP clients do not receive non-JSON-RPC log lines on stdout.
Vercel Deploy
Set MCP_API_KEY during setup and connect clients to /api/mcp/<surface>.
Passthrough tools are disabled unless both the surface declares the tool and the deployment enables it. For example, a surface with passthrough_tools: [write_content] exposes the writer only when MCP_ENABLE_WRITE_CONTENT=true and AI_GATEWAY_API_KEY is configured.
Security
Production deployments should use Authorization: Bearer <MCP_API_KEY>. Query-string API keys are an opt-in compatibility fallback only. This template does not implement OAuth; see Security before claiming OAuth compliance in a downstream deployment.
Example content is synthetic fixture content. Keep private memory, secrets, provider identifiers, and personal deployment domains in downstream forks.
Operational endpoints:
GET /api/healthGET /api/build-info
Documentation
- Architecture
- Content model
- MCP tool reference
- Vercel deployment
- Security
- MCP Inspector validation
- MCP client examples
- Provider integrations
- Passthrough tools
- Provider/platform validation notes
- Observability and logging
- Sync sources
- Downstream deployments
- Deployment model decision
- Storage adapter notes
- Operations runbook
- Registry readiness
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
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.