diff --git a/apps/web/content/blog/1password-vs-bitwarden-mcp-server.mdx b/apps/web/content/blog/1password-vs-bitwarden-mcp-server.mdx new file mode 100644 index 0000000..505a240 --- /dev/null +++ b/apps/web/content/blog/1password-vs-bitwarden-mcp-server.mdx @@ -0,0 +1,88 @@ +--- +title: "1Password vs Bitwarden MCP Server: Full Comparison" +description: "1Password and Bitwarden both ship MCP servers for AI agents, but their auth models solve the trust problem in opposite ways. Here's the real comparison." +excerpt: "1Password's MCP server never lets an agent see a secret value; Bitwarden's does, through a scriptable CLI. That one design choice shapes everything else, from auth model to automation friction to which server actually fits an unattended agent job." +date: "2026-08-11" +updatedAt: "2026-08-11" +author: "Gus Marquez" +authorUrl: "https://www.linkedin.com/in/gustavoamarquez" +authorSameAs: + - "https://www.linkedin.com/in/gustavoamarquez" +tags: + - "mcp" + - "developer" + - "security" + - "secrets-management" + - "password-manager" +category: "security" +focusKeyword: "1password vs bitwarden mcp server" +topic_source: "new_topic" +draft: false +cornerstone: false +faqItems: + - question: "Can I run both 1Password and Bitwarden's MCP servers in the same agent workflow?" + answer: "Yes, technically, though most teams standardize on one vault as their system of record. Running both just means a workflow needs credentials for two separate vendors, which widens what a compromised session could reach without adding real capability." + - question: "Does Bitwarden's MCP server expose Vaultwarden self-hosted instances too?" + answer: "The official Bitwarden MCP server targets Bitwarden's own cloud vault. MCPFind's directory separately indexes an unofficial community server that covers both Bitwarden and self-hosted Vaultwarden instances, useful if you are not on Bitwarden's hosted product." + - question: "Which server is safer for an agent that only needs to read secrets, not write them?" + answer: "1Password's, by design. Since it never returns a secret value to the model, a compromised or misbehaving agent session cannot exfiltrate a credential through the MCP layer even in a read-only workflow. Bitwarden's CLI tools return real values on every retrieval call, so scoping matters more there." + - question: "Do either 1Password or Bitwarden charge extra for their MCP server?" + answer: "No. Both connect to your existing subscription. 1Password's server ships as a Labs feature inside the desktop app you already have, and Bitwarden's is a free, open-source add-on to an existing Bitwarden account or organization." +--- + +1Password and Bitwarden both shipped official MCP servers this year. Same problem, both of them: let an AI agent reach a password vault without you pasting credentials into a chat window. On how much the agent should actually see, they disagree completely. 1Password's server never returns a secret value to the model. Bitwarden's does, through a CLI session token an agent can call directly. That one design choice ripples out into authentication, automation friction, and which of the two you'd actually trust with an unattended overnight job. Here's the comparison, past the marketing copy. + +| Server | Auth Model | Secret Exposure | Open Source | Best For | +|---|---|---|---|---| +| 1Password MCP Server | Desktop-app approval per request | Never exposed to model | No (bundled in app) | Interactive dev work, zero-trust posture | +| Bitwarden MCP Server | CLI session token + API credentials | Exposed via CLI tool results | Yes (GPL-3.0, ~231 stars) | Scheduled or unattended automation, org admin | + +## How Do 1Password and Bitwarden's MCP Servers Handle Secret Exposure Differently? + +1Password's MCP server never lets the model see a secret. Values get injected at runtime for an authorized process, and the agent only ever sees a variable name. Bitwarden goes the other way. Its CLI tools return the actual password, TOTP code, or note text as a tool-call result, so the real value lands in the agent's context window and stays there. + +The split comes down to which failure mode each vendor decided to design against. 1Password assumes the model itself is the risk: it can't leak what it never received. Bitwarden puts the trust boundary around the process running the CLI instead, and its tool list runs a lot longer as a result. An agent can unlock and lock the vault, retrieve items, create and edit logins, notes, and cards, generate new passwords, and spin up one-time Bitwarden Sends. 1Password's Environments-variable scope covers far less than that. Every Bitwarden retrieval call returns a real value into the context window. + +## Which MCP Server Has a Better Authentication Model for Unattended Agents? + +For work that runs without a person watching, Bitwarden's model is the practical one. You generate a CLI session token once with `bw unlock --raw`, hand it to a scheduled job, and it keeps working until it expires. 1Password wants a live desktop-app approval for every single access, tied to a person sitting at the machine. + +The rest of each product follows from there. 1Password's model kills any workflow that has to run while nobody is watching, because a batch of agents pulling secrets overnight never gets past the approval prompt. Bitwarden's org-admin surface reaches further still: member and group management, collection permissions, policy configuration, and audit logs, all through API credentials (`BW_CLIENT_ID`/`BW_CLIENT_SECRET`) that need no human in the loop at all. If a developer is sitting at the machine to click approve, 1Password's friction is cheap. + +## Is Either 1Password or Bitwarden's MCP Server Open Source? + +Bitwarden's is. GPL-3.0, published at github.com/bitwarden/mcp-server, sitting at roughly 231 GitHub stars as of this writing. Anyone can read the exact code deciding what an agent can and can't touch. 1Password ships its server bundled inside the desktop app itself, through Settings, Labs, MCP Server. No standalone repository to audit. + +For a security tool that gap matters more than it would in most other MCP categories. An open license is no guarantee of good code, though it does mean the auth logic handling your vault access isn't a black box you have to take on the vendor's word alone. Bitwarden's own README carries a caveat worth reading first: the server is meant for local use only and should never sit exposed on a public network. Open source doesn't mean safe to host remotely. Neither server holds the secrets anyway. Both sit in front of a vault that's still gated by the vendor's existing product security. + +Got a security review process for new tooling? The review looks different depending on which one you pick. Bitwarden hands an auditor a real repository to read line by line. 1Password hands that auditor vendor documentation and a track record, because the client-side code sits inside a closed desktop application. + +## What Can Bitwarden's MCP Server Do That 1Password's Can't? + +Organization administration, mostly. Bitwarden's server splits in two: CLI-based vault operations for individual use, and API-based admin tools scoped to running an entire Bitwarden organization. An agent can manage members and groups, configure collection permissions, pull audit logs, handle subscription details, and run bulk imports. 1Password's server touches none of that. + +1Password's Environments-focused design was never trying to compete there. It manages variables for a coding agent to consume as configuration, not an organization's whole identity and access posture. That's the right shape if what you actually need is narrowly-scoped runtime secrets with zero model exposure. Tool count isn't really the axis that separates these two. If your workflow needs an admin console an agent can drive, Bitwarden is the only one of the pair that has one. + +## Where Do These Servers Sit in MCPFind's Security Category? + +Neither official server has a healthy listing in MCPFind's [security category](/categories/security). The category now spans 1,016 indexed servers at a 0.08 average star count, which means most entries are thin, unmaintained, or built by one developer over a weekend. That count was 797 as recently as early August, so the category is growing fast even while quality stays flat. The two vendor-official tools compared here sit well outside that pattern. + +Community wrappers do exist for both products. [io-github-gambadio-onepassword-agent-mcp](/servers/io-github-gambadio-onepassword-agent-mcp) offers local access to approved 1Password fields. [io-github-aiwerk-mcp-server-vault](/servers/io-github-aiwerk-mcp-server-vault) covers Bitwarden and Vaultwarden, listing tagged items and one-time Sends. Both sit at zero stars. For the wider access-control picture, there's MCPFind's [security deep dive](/blog/mcp-server-security-deep-dive-permissions), the [OAuth and API key guide](/blog/mcp-server-authentication-oauth-api-keys), the existing [1Password setup guide](/blog/1password-mcp-server-secrets-management), and the [Auth0 vs Okta comparison](/blog/auth0-vs-okta-mcp-servers). New to MCP entirely? Start with the [cornerstone guide](/blog/what-is-mcp) first. + +## Frequently Asked Questions + +### Can I run both 1Password and Bitwarden's MCP servers in the same agent workflow? + +Yes, technically. Most teams standardize on one vault as their system of record though. Running both means a workflow carries credentials for two separate vendors, so you get a wider blast radius on a compromised session and no real capability in return. + +### Does Bitwarden's MCP server expose Vaultwarden self-hosted instances too? + +The official Bitwarden MCP server targets Bitwarden's own cloud vault. MCPFind's directory separately indexes an unofficial community server covering both Bitwarden and self-hosted Vaultwarden instances. Useful if you aren't on Bitwarden's hosted product. + +### Which server is safer for an agent that only needs to read secrets, not write them? + +1Password's, by design. It never returns a secret value to the model, so a compromised or misbehaving agent session can't exfiltrate a credential through the MCP layer even in a read-only workflow. Bitwarden's CLI tools return real values on every retrieval call. Scoping matters more there. + +### Do either 1Password or Bitwarden charge extra for their MCP server? + +No. Both connect to your existing subscription. 1Password's ships as a Labs feature inside the desktop app you already have. Bitwarden's is a free, open-source add-on to an existing Bitwarden account or organization. diff --git a/apps/web/content/blog/clickhouse-mcp-server-analytics-olap-guide.mdx b/apps/web/content/blog/clickhouse-mcp-server-analytics-olap-guide.mdx new file mode 100644 index 0000000..4a62cbb --- /dev/null +++ b/apps/web/content/blog/clickhouse-mcp-server-analytics-olap-guide.mdx @@ -0,0 +1,103 @@ +--- +title: "ClickHouse MCP Server Guide for Analytics and OLAP" +description: "ClickHouse's official MCP server connects AI agents to analytics and OLAP workloads with read-only defaults. Here's the setup, the tools, and the auth model." +excerpt: "ClickHouse ships a first-party MCP server that queries both ClickHouse Cloud and self-hosted instances, read-only by default, with writes gated behind explicit env vars. This guide covers install, auth, tool names, and where it fits next to Snowflake." +date: "2026-08-13" +updatedAt: "2026-08-13" +author: "Adam Bush" +authorUrl: "https://www.linkedin.com/in/adam-bush/" +authorSameAs: + - "https://www.linkedin.com/in/adam-bush/" +tags: + - "mcp" + - "developer" + - "databases" + - "analytics" + - "olap" +category: "databases" +focusKeyword: "clickhouse mcp server for analytics and olap workloads" +topic_source: "new_topic" +draft: false +cornerstone: false +faqItems: + - question: "Can ClickHouse's MCP server write or modify data, not just query it?" + answer: "Only if you turn it on. Writes are disabled by default (CLICKHOUSE_ALLOW_WRITE_ACCESS=false), and destructive operations like DROP or TRUNCATE need a second, separate flag (CLICKHOUSE_ALLOW_DROP=true) on top of that. An agent cannot touch your data by accident with default settings." + - question: "Does the ClickHouse MCP server work with the public ClickHouse SQL playground?" + answer: "Yes. The official README includes an example configuration pointed at the public sql-clickhouse.clickhouse.com playground, which is a useful way to try the server before connecting it to a real database." + - question: "Do I need a ClickHouse Cloud account to use this MCP server?" + answer: "No. The server connects to either ClickHouse Cloud or a self-hosted ClickHouse instance through the same set of environment variables, so a local or on-prem cluster works exactly the same way a cloud one does." + - question: "Is there an embedded, file-based option if I don't want to run a full ClickHouse server?" + answer: "Yes, through the chdb extra. Installing with pip install 'mcp-clickhouse[chdb]' adds a run_chdb_select_query tool backed by chDB, ClickHouse's embedded engine, so you can query local files without standing up a server at all." +--- + +Analytics databases showed up late to MCP. Transactional ones got there first, and ClickHouse's official server closes that gap with a scope that's narrower and more deliberate than most database integrations bother with. Read-only by default. It talks to ClickHouse Cloud and self-hosted clusters through the same environment variables, and there's an embedded option for querying local files without running a server at all. What follows: the real install command, the actual tool names, how auth works, and one genuine discrepancy between ClickHouse's own docs page and its GitHub README that you'll want to know about before you configure anything. + +## What Is the ClickHouse MCP Server and Who Publishes It? + +This one's first-party. ClickHouse publishes it at `github.com/ClickHouse/mcp-clickhouse` and links to it straight from their own documentation. Apache-2.0 licensed, roughly 850 GitHub stars as of this writing, which counts for something in a databases category where the average listed server barely clears single digits. + +It's built for analytics and OLAP work, not general-purpose database administration. The tool surface covers schema exploration and query execution against ClickHouse, and then there's a separate path for chDB, ClickHouse's embedded query engine, which lets you run analytical queries against local files with no running server at all. That split is useful if your workflow spans a live cluster and ad hoc file analysis both. You cover both cases from a single install with different tools, so there's no second integration sitting around that you also have to keep alive. + +## How Do You Install and Connect the ClickHouse MCP Server? + +There are two install paths. The recommended one leans on `uv`, which sorts out the Python environment for you and skips the separate virtual environment step: + +```json +{ + "mcpServers": { + "clickhouse": { + "command": "uv", + "args": ["run", "--with", "mcp-clickhouse", "--python", "3.10", "mcp-clickhouse"], + "env": { + "CLICKHOUSE_HOST": "your-instance.clickhouse.cloud", + "CLICKHOUSE_USER": "default", + "CLICKHOUSE_PASSWORD": "your-password" + } + } + } +} +``` + +The other option is a plain `pip install mcp-clickhouse`. Add `pip install "mcp-clickhouse[chdb]"` if you want the embedded chDB tool too. A handful of optional environment variables let you tune past the basics: `CLICKHOUSE_PORT`, `CLICKHOUSE_SECURE`, `CLICKHOUSE_VERIFY`, `CLICKHOUSE_CONNECT_TIMEOUT`, `CLICKHOUSE_ROLE`, and `CLICKHOUSE_DATABASE` all narrow or adjust the connection without you touching any code. The README even ships an example pointed at the public `sql-clickhouse.clickhouse.com` playground, which is a fast way to test the whole setup before you wire in real credentials. + +## What Tools Does the ClickHouse MCP Server Expose? + +The GitHub README lists three core tools: `run_query`, `list_databases`, and `list_tables`. Add `run_chdb_select_query` for the embedded chDB path, plus a `/health` endpoint for monitoring. The surface is deliberately narrow: you find out what's there, then you query it, and that's the whole loop. + +The two published sources disagree on one tool name. ClickHouse's own documentation page calls the query tool `run_select_query`. The GitHub README, which is the more actively maintained source, calls it `run_query`. So if your client shows a tool name you weren't expecting after connecting, that mismatch between the docs page and the current code is the likely reason, not a broken install. Go with the README until ClickHouse reconciles the two. + +## What Kinds of Questions Can an Agent Answer With the ClickHouse MCP Server? + +Anything that maps to a read-only analytical query against tables it can already see. A typical exchange opens with `list_databases`, moves to `list_tables` once the agent settles on one, then lands on `run_query` for the actual SELECT statement, usually with a GROUP BY or a window function in it, because that's the kind of question OLAP databases exist to answer quickly. Ask something like "which product category drove the most revenue last month" and the agent walks that same three-step path before coming back with a real answer instead of a guess. + +The small tool set works in your favor here. An agent holding only `run_query`, `list_databases`, and `list_tables` can't wander into schema changes or administrative territory by accident, because those operations aren't exposed as tools at all. Teams pointing agents at production analytics data on a recurring basis get more real safety out of that constraint than out of any prompt instruction telling the model to be careful. + +## Is the ClickHouse MCP Server Safe to Run Against a Production Cluster? + +Reasonably, yes, as long as you leave the defaults alone. Queries run read-only unless you go set `CLICKHOUSE_ALLOW_WRITE_ACCESS=true` yourself. Even then, destructive statements like DROP or TRUNCATE need a second, separate flag on top of that one, `CLICKHOUSE_ALLOW_DROP=true`. An agent connected with default settings simply cannot alter your data. That restriction lives in the server's own configuration defaults, so it holds whether or not anyone on your team remembers to be careful. + +The two-flag structure is meaningfully more conservative than what a lot of database MCP servers ship with, where read and write access tend to sit behind the same single credential. For a production analytics cluster, the practical move is to connect with a scoped, read-only ClickHouse user and leave the write flags off. The database-level permission then holds regardless of what the server config says, or who copies that config where. + +## How Does ClickHouse Compare to Other Databases in MCPFind's Directory? + +MCPFind's [databases category](/categories/databases) indexes 576 servers at a 7.68 average star count, well above the near-zero averages you see in newer categories. ClickHouse's own server would land near the top of that field if it were indexed today. It isn't in MCPFind's directory yet as an official listing. The closest match currently indexed is [DB Connect MCP](/servers/io-github-yugui923-db-connect-mcp), an unofficial multi-database server that reaches PostgreSQL, MySQL, and ClickHouse through a single connector. So the gap here is in directory coverage, not in the server, which is in fine shape. + +For teams already running [Snowflake](/blog/snowflake-mcp-server-enterprise-data), the difference that matters is concurrency. ClickHouse's own engineering team cites support for over 1,000 concurrent queries against roughly eight per warehouse on Snowflake. Real architectural gap, and worth weighing if your agent workloads fire off many queries in parallel. If Postgres-flavored analytics is more your speed, MCPFind's guides to [Neon](/blog/neon-mcp-server-serverless-postgres), [self-hosted Supabase](/blog/self-hosted-supabase-mcp-server), [Redis](/blog/redis-mcp-server-caching-guide), and [MySQL](/blog/mysql-mcp-server-setup-guide) cover the rest of the databases cluster, with pretty varied setup patterns between them. The [cornerstone guide](/blog/what-is-mcp) covers MCP itself, which is the place to start if you're wiring a database server into a live workflow for the first time. + +## Frequently Asked Questions + +### Can ClickHouse's MCP server write or modify data, not just query it? + +Only if you turn it on. Writes are off by default (`CLICKHOUSE_ALLOW_WRITE_ACCESS=false`), and destructive operations like DROP or TRUNCATE need a second, separate flag (`CLICKHOUSE_ALLOW_DROP=true`) on top of that. With default settings, an agent can't touch your data by accident. + +### Does the ClickHouse MCP server work with the public ClickHouse SQL playground? + +Yes. The official README includes an example configuration pointed at the public `sql-clickhouse.clickhouse.com` playground, which is a handy way to try the server before you connect it to a real database. + +### Do I need a ClickHouse Cloud account to use this MCP server? + +No. The server connects to either ClickHouse Cloud or a self-hosted ClickHouse instance through the same set of environment variables, so a local or on-prem cluster works exactly the way a cloud one does. + +### Is there an embedded, file-based option if I don't want to run a full ClickHouse server? + +Yes, through the chdb extra. Install with `pip install "mcp-clickhouse[chdb]"` and you get a `run_chdb_select_query` tool backed by chDB, ClickHouse's embedded engine, so you can query local files without standing up a server at all. diff --git a/apps/web/content/blog/pinterest-microsoft-advertising-mcp-servers.mdx b/apps/web/content/blog/pinterest-microsoft-advertising-mcp-servers.mdx new file mode 100644 index 0000000..8d95683 --- /dev/null +++ b/apps/web/content/blog/pinterest-microsoft-advertising-mcp-servers.mdx @@ -0,0 +1,85 @@ +--- +title: "Pinterest vs Microsoft Advertising MCP Servers" +description: "Pinterest and Microsoft Advertising both launched official MCP servers on June 17, 2026. Here's what each one actually exposes, and how to get access to it." +excerpt: "Two ad platforms shipped MCP servers on the same day, one week ahead of Cannes Lions. Pinterest's is an invite-only alpha; Microsoft's is a broadly available pilot. Neither is indexed in MCPFind's directory yet. Here's the real difference between them." +date: "2026-08-12" +updatedAt: "2026-08-12" +author: "Adam Bush" +authorUrl: "https://www.linkedin.com/in/adam-bush/" +authorSameAs: + - "https://www.linkedin.com/in/adam-bush/" +tags: + - "mcp" + - "beginner" + - "advertising" + - "marketing" + - "pinterest" +category: "other" +focusKeyword: "pinterest and microsoft advertising mcp servers compared" +topic_source: "new_topic" +draft: false +cornerstone: false +faqItems: + - question: "Can anyone use Pinterest's new MCP server right now?" + answer: "Not yet. Pinterest launched it as an invite-only alpha with a named set of agency partners: PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom's Jump450. There is no public signup listed as of this writing." + - question: "Is Microsoft Advertising's MCP server free to use?" + answer: "Yes, Microsoft's own product page describes it as free to connect. You still need an active Microsoft Advertising account with campaigns to query, the MCP server itself does not carry a separate fee." + - question: "Can either MCP server create or edit ad campaigns, or only read data?" + answer: "Both are read-only at launch. Pinterest's exposes campaign, analytics, and keyword insight data; Microsoft's exposes campaign, ad group, ad, and keyword data plus performance metrics. Neither one lets an agent push changes back to a live campaign yet." + - question: "Why did Pinterest and Microsoft Advertising launch MCP servers on the same day?" + answer: "Both launches landed June 17, 2026, roughly a week before Cannes Lions, the advertising industry's biggest annual event. That timing is common for ad-tech announcements meant to set the agenda going into the conference, though neither company has stated the overlap was coordinated." +--- + +Two competing ad platforms shipped official MCP servers on the exact same day. June 17, 2026. Pinterest and Microsoft Advertising both went live roughly a week ahead of Cannes Lions, and both now hand AI agents structured access to campaign and performance data instead of a dashboard export you have to babysit. The date matches and the protocol matches, but the two products barely resemble each other. One is a closed alpha with six named partners. The other, anybody with an account can join. Here's what each server actually exposes, who can reach it today, and where the two of them land next to the third-party ad tools MCPFind already indexes. + +## What Do Pinterest and Microsoft Advertising's New MCP Servers Actually Do? + +Both hand an AI agent structured, read-only access to your advertising account data. No spreadsheet export, no dashboard screenshot pasted into a chat window. Pinterest's server surfaces campaign, analytics, and keyword insight data, and it also carries the taste, trend, and intent signals that Pinterest's whole ad platform is built around. Microsoft Advertising's covers more familiar PPC ground: campaigns, ad groups, ads, and keywords, plus the performance metrics you'd expect. Spend, clicks, impressions, click-through rate, conversions, return on ad spend. + +Where both stop short is on writes. Your agent can pull a performance report or summarize how a campaign's keyword mix is doing, but it can't pause that campaign or nudge a bid through the MCP server. + +## How Does Pinterest's MCP Server Differ From Microsoft Advertising's? + +Start with who can actually get in, because that's where the two diverge hardest. Pinterest went with an invite-only alpha and named exactly six agency partners: PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom's Jump450. Microsoft Advertising went the other direction with an open pilot, a meaningfully broader rollout that any advertiser with an account can join. + +Then there's the data itself. Pinterest's platform runs on visual discovery, so its server pushes taste graph and intent signals forward alongside the standard campaign metrics, and that's data with no clean equivalent over on a traditional search-ads platform. Microsoft Advertising stays closer to conventional PPC reporting. Campaign and keyword performance, basically the shape any agent trained on Google Ads or Meta Ads reporting already recognizes on sight. Running both platforms at your agency? Plan on the Microsoft integration being live and useful long before Pinterest's alpha opens up further. + +## Which AI Tools Can Connect to These Ad-Platform MCP Servers? + +Microsoft Advertising supports the widest client list at launch: M365 Copilot through Copilot Studio, Claude desktop, ChatGPT desktop, and any other MCP-compatible client. Because the server is a standard MCP endpoint rather than something bolted to one vendor's tooling, the client list is really just whatever speaks MCP, and that wide-open client story lines up with the wide-open access model by design. + +Pinterest hasn't published a general client list. With the alpha scoped to six named partners, client support so far is whatever those six agencies happen to run internally, and there's no public integration guide to read. If you're not one of the six, you don't have a documented path to connect any client to Pinterest's MCP server, alpha access or not. That'll probably change as the program widens, but nothing in Pinterest's own announcement puts a date on it. + +## Why Are Ad Platforms Racing to Launch Official MCP Servers Now? + +Neither company invented this move. Developers were building third-party MCP connectors for ad platforms long before either vendor shipped anything official, because agencies wanted agent access to campaign data badly enough to just go build it themselves. MCPFind's directory already carries community versions for [Reddit Ads](/servers/io-github-mkerchenski-reddit-ads-mcp), [Meta Ads](/servers/io-github-mike25app-scaleforge-mcp-meta-ads), and [LinkedIn Ads](/servers/io-github-pipeworx-io-linkedin-ads). None of them affiliated with the platforms they touch. + +Enough unofficial tooling piles up in a category, the platform notices, and an official launch follows on the vendor's own terms: tighter scopes, read-only defaults, and a rollout the platform controls rather than one shaped by whatever a community project decided to expose. Pinterest and Microsoft both fit that shape, and more ad platforms probably will over the next year. + +## Are Pinterest and Microsoft's Official Ad MCP Servers in MCPFind's Directory Yet? + +Neither one is indexed yet. The directory tracks third-party ad-platform servers today, and these two official launches are simply too new, plus too access-restricted in Pinterest's case, to have landed in the index. Indexing takes something public to point at: a repo, or a documented endpoint with connection instructions. Pinterest's alpha has neither outside its six partners. + +The related third-party tooling already in there does tell you how crowded this corner is getting. [Pinterest Ads MCP Server](/servers/io-github-cesteral-pinterest-mcp) is an unofficial, community-built option covering Pinterest ad CRUD operations. A separate [multi-platform ad connector](/servers/io-github-opusgrowth-ads-mcp) reaches Google, Microsoft, TikTok, and LinkedIn Ads across roughly 233 tools. Neither is affiliated with Pinterest or Microsoft. And MCPFind's [social category](/categories/social) holds 262 servers at a 0.11 average star count, thin by any measure, so an official vendor listing from either company would stand out considerably once it lands. Both of those community entries carry zero recorded GitHub stars in MCPFind's index right now, which tracks with how young and narrowly scoped the whole official ad-platform-MCP category still is. + +## Frequently Asked Questions + +### Can anyone use Pinterest's new MCP server right now? + +Only the six named alpha partners, for now. Pinterest launched it as an invite-only alpha: PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom's Jump450. No public signup is listed as of this writing. + +### Is Microsoft Advertising's MCP server free to use? + +Yes. Microsoft's own product page describes it as free to connect. You'll still need an active Microsoft Advertising account with campaigns in it to query, but the MCP server itself doesn't carry a separate fee. + +### Can either MCP server create or edit ad campaigns, or only read data? + +Both servers ship read-only at launch. Pinterest's side gives you campaign, analytics, and keyword insight data, while Microsoft's covers campaign, ad group, ad, and keyword data plus the performance metrics on top of that. No agent is pushing a change back into a live campaign yet. + +### Why did Pinterest and Microsoft Advertising launch MCP servers on the same day? + +Both landed June 17, 2026, roughly a week before Cannes Lions, the advertising industry's biggest annual event. Ad-tech announcements cluster there on purpose, since the goal is to set the agenda going into the conference. Neither company has stated the overlap was coordinated. + +New to how any of this works under the hood? Start with MCPFind's [cornerstone guide to MCP](/blog/what-is-mcp), before you wire an advertising server into anything. + +A few others worth your time. [Docker MCP Catalog vs MCP Directory](/blog/docker-mcp-catalog-toolkit-guide) gets at how the ecosystem discovers new servers like these two in the first place, and if you're building the agent side instead of just querying it, [OpenAI Agents SDK vs MCP](/blog/openai-agents-sdk-vs-mcp) is the one to read. Vendor-hosted launch versus community project? That comparison lives in [Open-Source vs Commercial MCP Servers](/blog/open-source-vs-commercial-mcp-servers-2026). Pinterest and Microsoft both ship theirs as hosted endpoints, so [Remote MCP Servers vs Local](/blog/remote-vs-local-mcp-server) applies here too. And [Using MCP Servers with Gemini CLI](/blog/mcp-servers-gemini-cli-setup-guide) walks a client-side setup on a totally different tool, same pattern underneath.