MCP server stacks
Don't pick servers one at a time — start from the job. Each stack is a curated bundle for a common build, with one-click merged config and a tool-budget check so you don't blow past Cursor's tool limit.
Retrieve, embed, and answer with sources — the retrieval stack.
3 servers
Repo access, local files, and current docs — the coding stack.
3 servers
Schema inspection + SQL across Postgres, ClickHouse, BigQuery.
3 servers
Search + scrape + synthesize, with sources.
3 servers
Repos, clusters, and monitoring in one agent.
3 servers
Correlate logs, metrics, and errors during incidents.
3 servers
Tickets, team chat, and a knowledge base.
3 servers
Generate images, audio, and source material.
3 servers
Drive real browsers locally or in the cloud.
3 servers
Payments, banking data, and market data — sandbox first.
3 servers
Issues, repos, and product analytics in one loop.
3 servers
Notion + Obsidian + docs as the AI's long-term memory.
3 servers
Prefer to browse by single capability? See MCP servers by capability .