The Databricks MCP server, built by Databricks, provides managed MCP servers that Databricks runs inside your own workspace, which is why there is nothing to install for the common case — you point an MCP client at a workspace URL and authenticate on behalf of the user with an OAuth scope. It is officially maintained and best for Database.
by Databricks
About
managed MCP servers that Databricks runs inside your own workspace, which is why there is nothing to install for the common case — you point an MCP client at a workspace URL and authenticate on behalf of the user with an OAuth scope. The hosted endpoints are /api/2.0/mcp/genie and /api/2.0/mcp/genie/{genie_space_id} for natural-language analytics against a Genie space (scope genie, read-only), /api/2.0/mcp/ai-search/{catalog}/{schema}/{index_name} for retrieval over AI Search indexes (scope ai-search; the older /api/2.0/mcp/vector-search/ path still resolves after the rename), /api/2.0/mcp/sql for running AI-generated SQL including writes (scope sql), and /api/2.0/mcp/functions/{catalog}/{schema}/{function_name} for executing SQL tools stored as Unity Catalog functions (scope unity-catalog). All of them prefix the workspace hostname, and because permissions are enforced on-behalf-of the calling user, an agent can only reach the data and tools that user could reach — the reason most teams choose these over a self-hosted server. For self-hosting there is databrickslabs/mcp, an experimental repo whose Unity Catalog server runs locally over stdio via uv and infers its tool list at startup from the UC functions, vector search indexes and Genie spaces in the schema you point it at; it can also be deployed onto Databricks Apps with databricks bundle or databricks apps, where it is reached over Streamable HTTP at a URL that must end in /api/mcp/ including the trailing slash. Two caveats worth knowing before you build on it: Databricks has marked that Unity Catalog server deprecated in favour of the managed servers, and the developer-tools server in the same repo is still labelled under construction and not usable. The separate databricks-mcp package on PyPI is a client-side helper library from databricks/databricks-ai-bridge (OAuth provider and MCP client plumbing for notebooks, Model Serving and local runs), not a server.
Trust verdict
How grades are computed →Grade B (81/100, solid) from 1 measured signal, based on repository evidence. Only one signal stands behind it, so treat the grade as provisional.
What was measured
- Repository maintenance55/100 · weight 20
The last push or release is between six and eighteen months old. — last push 2025-10-10.
- Source verification100/100 · weight 25
The repository URL was confirmed to resolve against the live GitHub API and is not archived.
- Provenance90/100 · weight 10
Published and maintained by the vendor of the service it connects to, rather than by a third party.
- Listing ↔ repository match70/100 · weight 5
The linked repository databrickslabs/mcp was confirmed to exist, but its name does not obviously correspond to this listing. Worth opening the repo before you trust the mapping.
What could not be measured
These contributed nothing to the score — not a penalty, not a zero. They are why the confidence reads the way it does.
- Live MCP handshakeunknown
No remote endpoint to handshake — this server installs and runs locally over stdio, so there is nothing to probe from the outside.
- Measured uptimeunknown
No probe history recorded for this server yet.
- Tool-schema stabilityunknown
Drift is a difference between two successive checks, and this server has none recorded.
Categories
Frequently Asked Questions
What is Databricks MCP Server?
Who built Databricks MCP Server?
Is Databricks MCP Server free?
How do I install Databricks MCP Server?
What does Databricks MCP Server integrate with?
Related Guides
Repo Health
Local/stdio install — runs on your machine, so there is no remote endpoint to verify live. Trust signal below is from the source repo.
- Last commit
- 11mo ago
- Install
- pip
Quick Info
- Install Type
- pip
- Author
- Databricks
- Categories
- 3
- Integrations
- 5
Related Servers
Memory
Knowledge graph-based persistent memory system. Store and retrieve contextual information.
Sequential Thinking MCP Server
a single structured-reasoning tool that lets a model plan, revise and branch its own chain of thought instead of answering in one shot. Published by Anthropic as part of the official modelcontextprotocol/servers monorepo (89,000+ stars, actively maintained), it exposes exactly one tool — sequential_thinking — and that tool is the whole product. Each call carries a `thought` string plus bookkeeping fields: `thoughtNumber`, `totalThoughts`, and `nextThoughtNeeded`, which the model flips to false when it is done. The interesting fields are the optional ones. `isRevision` and `revisesThought` let the model go back and correct an earlier step rather than plowing ahead on a bad assumption; `branchFromThought` and `branchId` let it fork into an alternative line of reasoning and carry both forward; `needsMoreThoughts` lets it extend past its own original estimate when a problem turns out to be deeper than it looked. In practice you never call the tool by hand. You connect the server to an MCP host and ask a question that deserves more than one pass — plan a PostgreSQL 14 to 16 migration and revise if downtime exceeds five minutes, work out why a deploy only fails in production, compare three architectures and branch when an assumption breaks. You can tell it is working when the host inspector shows repeated sequential_thinking calls with a rising `thoughtNumber` rather than a single response. Install with `npx -y @modelcontextprotocol/server-sequential-thinking` — note the hyphenated package name, which differs from both the `sequentialthinking` directory in the repo and the Docker image `mcp/sequentialthinking`, a mismatch that breaks a lot of copied configs. A Docker image is published alongside the npm package, and the README carries one-click VS Code install buttons for both transports. Set `DISABLE_THOUGHT_LOGGING=true` if you do not want every thought written to the server log.
Exa MCP Server
Exa's official MCP server connects AI assistants to a search engine purpose-built for AI, using neural embeddings to match on meaning rather than keywords so agents get clean, ready-to-use content instead of a page of blue links to re-parse. The default tool set covers web_search_exa for quick topical lookups and web_search_advanced_exa for full control over domains, date ranges, and content filters, plus specialized tools for code_search (searching real-world code and GitHub), company_research (building company profiles, competitor lists, and financials), crawling/web_fetch (pulling clean content from a specific URL), people_search and linkedin_search (public professional-profile lookups), and deep_researcher_start/check for long-running multi-step research tasks backed by Exa's Research API. The server is hosted at https://mcp.exa.ai/mcp — no local process to run — and connects via one-line setup in Cursor, VS Code, Claude Code, Claude Desktop (available as a native Connector), Codex, OpenCode, Windsurf, and Antigravity, authenticated with an EXA_API_KEY from the Exa dashboard. Tool exposure is tunable per client via a ?tools= query parameter on the endpoint URL, letting teams ship narrow, purpose-built configurations (e.g. company-research-only or LinkedIn-only agents) instead of exposing the full surface, and Exa ships ready-made Claude Skills/agent definitions for common patterns like company research and people search with built-in query-variation and token-isolation guidance.
MongoDB MCP Server
The official MongoDB MCP server, `mongodb-js/mongodb-mcp-server`, maintained by MongoDB. Searchers find it as the MongoDB MCP server or the Mongo MCP server; both names refer to this one project, and there is no separate short-form server. It is really two tool surfaces in one process. The database tools connect to any MongoDB deployment over a connection string in `MDB_MCP_CONNECTION_STRING` — `find`, `aggregate`, `explain`, `collection-schema`, `collection-indexes`, `create-index`, `count`, `export`, `mongodb-logs` and the write tools — or, if no connection string is set, the model calls the `connect` tool at runtime and passes the returned `connectionId` to later calls. The Atlas tools are a separate control plane: listing and creating clusters, database users, IP access-list entries, stream processing resources and the performance advisor. They authenticate with Atlas Service Account credentials (`MDB_MCP_API_CLIENT_ID` / `MDB_MCP_API_CLIENT_SECRET`), not with a connection string, and they simply do not register when those are unset — the usual cause of a 'the Atlas tools are missing' report. Two defaults are worth knowing before you point it at production. `readOnly` is false, so every write tool including `drop-database` is registered unless you pass `--readOnly`, which is why MongoDB's own README uses that flag in every example. And confirmation-required tools rely on MCP elicitation, so on a client that does not support elicitation they execute without prompting — `--readOnly` or `--disabledTools` is the actual boundary. Other guardrails: `indexCheck` rejects queries that would do a collection scan, server-side JavaScript (`$where`, `$function`, `$accumulator`) is disabled by default, results are capped at 100 documents and 16 MB per call, and `--dryRun` prints the resolved config and enabled tool list without starting the server. Requires Node 22.13.0 or later (Node 20 is deprecated); also published as the `mongodb/mongodb-mcp-server` Docker image. Runs over stdio by default, with an optional HTTP transport and a separate monitoring listener for `/health` and Prometheus `/metrics`.
Codex MCP Server
OpenAI Codex is a lightweight coding agent that runs in your terminal, and it sits on both sides of the Model Context Protocol — which is exactly the confusion the "codex mcp" search hides. As a client, Codex connects to other MCP servers you configure under `[mcp_servers]` in `~/.codex/config.toml`, and the `codex mcp` subcommand manages those configured launchers. As a server, Codex exposes its own experimental MCP interface: run `codex mcp-server` (alias `codex-mcp-server`) and another MCP client can drive a local Codex engine over stdio using JSON-RPC 2.0. That server interface is genuinely useful and genuinely experimental — the docs state it is subject to change without notice — and it publishes methods to manage threads, turns, accounts, config and approvals: v2 RPCs like `thread/start`, `thread/resume`, `thread/fork`, `turn/start`, `turn/steer`, `turn/interrupt`, `account/login/start`, `config/read`/`config/value/write`, `model/list`, plus v1 compatibility calls (`getConversationSummary`, `getAuthStatus`, `gitDiffToRemote`, fuzzy file search) and streaming `codex/event/*` notifications. Approvals flow the other way as server-to-client requests (`applyPatchApproval`, `execCommandApproval`), so a host UI can gate patch and exec actions. Install the CLI with `npm install -g @openai/codex` (or Homebrew), then start the server with `codex mcp-server`; a quick inspection UI is `npx @modelcontextprotocol/inspector codex mcp-server`. Built in Rust, Apache-2.0 licensed, and by far the most-starred entry in this space at 100,000+ stars. Use this when you want an AI client to orchestrate a real Codex coding session rather than reimplement one.
Sponsored
Better Stack
Free PlanGet alerted when your APIs, browser tests, payment pipelines, or MCP server dependencies go down. Used by 100K+ developers.
Start monitoring free →