Directory
MCP Servers for Agent Workspaces
Model Context Protocol is a standard way for an AI agent to discover and call tools outside its own process. An MCP server is one of those tools, and this is a curated directory of real ones you can connect to a coding agent today.
Filesystem and shell
Give an agent scoped access to local files and commands.
Read, write, edit, move, and search files inside directories you explicitly grant.
Run terminal commands, manage long-running processes, and make diff-based edits to local files.
Version control
Read and act on git history, issues, and pull requests.
Run git operations (status, diff, log, commit, branch) on a local repository.
Work with repositories, issues, pull requests, Actions, and code search over the GitHub API.
Token-based access to any GitLab instance: issues, merge requests, files, and pipelines.
Databases
Query and inspect a database with the agent's own permissions.
Create and manage Neon serverless Postgres projects and branches, and run SQL via the Neon API.
Run SQL against a Supabase project, inspect schema, manage branches, and generate types.
Read and write Redis data structures and vector or search indexes through tool calls.
Query and administer MongoDB databases and Atlas clusters, and run find and aggregate operations.
Read and write SQL, inspect schema, and analyze query performance for PostgreSQL.
Browsers and scraping
Drive a browser or pull structured content from the web.
Fetch a URL and convert its HTML to markdown for the model to read.
Drive a browser through Playwright's accessibility tree to navigate, click, type, and read pages.
Control Chrome over the DevTools Protocol for navigation, DOM and network inspection, and traces.
Scrape, crawl, and search websites, returning cleaned markdown or structured data.
Run cloud headless-browser sessions for navigation and extraction, driven with Stagehand.
Run Apify Store scrapers to extract data from sites, search results, maps, and social platforms.
Communication and project tools
Read and post to the tools a team already works in.
Read and write Notion pages, databases, and blocks through the Notion API.
Find, create, and update Linear issues, projects, and comments through Linear's hosted server.
Search and edit Jira issues and Confluence pages through Atlassian's hosted server.
Read and post Slack messages across channels and DMs, with history, and no bot app to install.
Send and read Discord messages and manage channels and servers via a bot token.
Cloud, deploy, and observability
Reach cloud accounts, deploys, errors, and payments.
A set of servers for Workers, DNS, R2, D1, KV, logs, and other Cloudflare account resources.
A suite of servers for AWS docs, CDK, Terraform, cost analysis, and individual services.
Connect agents to Azure resources, storage, databases, and monitoring via Azure APIs and CLI.
Manage Vercel projects, deployments, and runtime logs, and search the Vercel docs.
Create projects, builds, and deploys, and manage Netlify sites via the Netlify API.
Query Sentry issues, errors, and traces, and run Seer root-cause analysis.
Query Grafana dashboards, datasources, Prometheus and Loki, incidents, and alerts.
Query PostHog analytics, insights, feature flags, and error tracking.
Create and read customers, products, prices, invoices, and subscriptions, and search Stripe docs.
Knowledge and memory
Persist context and search it back through vectors or a graph.
Store and retrieve facts as a persistent local knowledge graph of entities and relations.
Store and semantically search documents and embeddings in a Qdrant collection.
Insert and retrieve objects from Weaviate collections using vector and hybrid search.
Search Pinecone docs and upsert or query vector data in Pinecone indexes.
Web search
Answer questions against the live web.
Run Brave web, image, news, video, and local queries via the Brave Search API.
Perform web search, content extraction, site mapping, and crawling via Tavily.
Run Exa web search and page-content retrieval for agents.
Run Kagi web search and related tools, including summarization, via the Kagi API.
Query the Perplexity API for web-grounded answers with citations.
Perform DuckDuckGo web search and fetch and parse result content, no API key.
Independently curated. Every server here was verified to exist, to resolve, and to be actively maintained at the time it was added. No listing is paid, sponsored, or ranked, and MoggingLabs is not affiliated with these projects. A link-checker keeps the list honest; anything that goes dark comes off.
Running these in one place
Every server above has to be registered in each agent CLI you use, and Claude Code, Codex, and Gemini each keep that config in a different file and format. Doing it by hand does not scale past a few servers. The guide to managing MCP servers walks through where each CLI reads its config and how to edit it safely.
MoggingLabs Workspace runs your agent CLIs in one terminal and is building a manager that registers a server once and writes it into every hosted CLI's format, with a backup and a preview before each write. It is not in a public build yet.