Research report · ai-agents
What Is MCP, and Which AI Agents Mention It?
We searched each listing's own attributes, key features and integrations for the text "MCP" on 27 August 2026: 70 of 94 mention it, 24 do not.
On this page
MCP mentions by category
Of the 94 AI agent product listings excluding the one agency listing, of the 95 published on 27 August 2026, the largest group is “Coding agents” at 20 (21%). Measured 2026-08-27.
Of the 94 AI agent products published in The Agents Index on 27 August 2026, 70 mention MCP support in their own recorded key features or integrations, and 24 do not. That is a text match over what our own research already says about each listing, not a verified claim about which tools actually implement the protocol.
Correction, 11 September 2026. This report was published as “What Is MCP, and Which AI Agents Actually Support It?”, with a lede reading “63 document support for the Model Context Protocol” and a Method section promising a count of listings with “an explicit MCP claim with a cited source.” That per-listing verification is not recorded anywhere a collector run can pin: our warehouse holds no field for it, so the 63 could never be re-derived, only re-typed. The only reading of this index that a fixed collector run does produce is simpler than what we published: whether the string “MCP” appears, case-insensitive, anywhere in a listing’s own recorded attributes, key features or integrations text. At collector run 221173 of 27 August 2026, that reading returns 70 of 94. This edition is restated to it, and the title, the lede, the category table, the roster of listings with no mention and the Method section below all move with it.
Almost every listing page in this space now advertises MCP support somewhere, usually with no explanation of what it buys a buyer. Below is the plain definition, the reason it matters when you are choosing an agent to build on, and the count from our own index.
What MCP actually is
The Model Context Protocol is an open standard, introduced by Anthropic on 2024-11-25 and open-sourced the same day, for connecting AI applications to external data and tools through one common interface instead of a custom integration per pair. Anthropic’s own framing is the clearest: before MCP, “every new data source requires its own custom implementation, making truly connected systems difficult to scale.” That is the N-times-M problem, where N AI applications each need a bespoke connector to M tools. MCP replaces the grid with one protocol both sides implement once.
The protocol’s documentation describes it as “a standardized way to connect AI applications to external systems,” and reaches for USB-C as the analogy: one interface that many devices and many hosts can both target, rather than a proprietary cable per pairing. An MCP server exposes a tool, database or workflow. An MCP client connects to any MCP server without glue code written for that specific pairing.
MCP is no longer an Anthropic convention. OpenAI documents MCP support in its own API, and MCP clients now ship inside Claude, ChatGPT, Visual Studio Code and Cursor. Once a server exists for a tool, any MCP-enabled agent can reach it.
Why MCP support changes what an integration costs you
If you are evaluating an agent to connect to your database, your internal APIs, your ticketing system or your CI pipeline, MCP support changes the shape of that work. Without it, “does this agent talk to my systems” is a bespoke engineering question, answered by whatever connectors the vendor happened to ship.
With it, the agent can reach any existing MCP server, including ones your own team writes once in-house, without a new custom integration per data source. That is a different buying question from “does it have an API.” A documented API tells you the agent can be scripted. MCP support tells you it plugs into a shared ecosystem of servers other people have already built.
Our index, counted on 27 August 2026
70 of the 94 AI agent products published in The Agents Index on 27 August 2026 mention MCP support in their own recorded key features or integrations. 24 of the 94 AI agent products published in The Agents Index show no mention of MCP in their own recorded key features or integrations, which is a narrower claim than saying they do not support it. The corpus also holds one agency listing, Agentix Labs, which sells engineering services rather than a product with an integration surface of its own, so it sits outside this count entirely.
Read 70 as a text match, not a market share and not a verified claim. We do not survey vendors, and this count does not confirm that a mention is accurate, current or complete: it confirms only that the word appears in the key features or integrations copy we have already researched and published for that listing. A tool that supports MCP but whose listing has not yet had that fact written into its key features or integrations counts against this number, not for it. A current, browsable list built the same way is maintained as a standing set: AI Agent Tools With MCP Support.
Sales and marketing agents mentions MCP least, and stays last
7 of the 8 categories in our index mention MCP on at least half their members; sales and marketing agents mention it on three of eight.
| Category | Mentioning MCP | In the index | Share |
|---|---|---|---|
| Agent frameworks | 14 | 14 | 100% |
| Research and data agents | 9 | 11 | 82% |
| Coding agents | 20 | 25 | 80% |
| Agent platforms | 12 | 16 | 75% |
| Support agents | 6 | 9 | 67% |
| Voice agents | 8 | 12 | 67% |
| Agent tools | 1 | 2 | 50% |
| Sales and marketing agents | 3 | 8 | 38% |
(Botsify, Sierra and ElevenLabs Conversational AI each belong to two categories and are counted once in each. All three mention MCP, so these rows sum to 97 members and 73 mentioning against 94 distinct tools and 70 distinct mentioning. The same convention applies in our deployment census.)
8 sales and marketing agents are published in the index, and 3 of them mention MCP in their own recorded key features or integrations. 3 of the 8 published sales and marketing agents mention MCP in their own recorded key features or integrations: 11x, Amplemarket and Lead Scorer. Amplemarket is new to this reading: an earlier edition of this page placed it among the listings with no MCP claim in our research at all, but its own recorded key features already say “MCP server for Claude and ChatGPT.” 11x’s own text is thin by comparison, a single “Custom MCP Integration” bullet on an enterprise pricing tier, with no docs and no stated client or server direction. Lead Scorer’s mention is the least ambiguous of the three: its key features and integrations both describe a single OAuth 2.1 endpoint at mcp.lead-scorer.com/mcp exposing 112 tools, documented on its own MCP page. Artisan, Qualified, Rox, Unify and Warmly are the other five in the category, and none mentions MCP in its recorded text. The likely reason is commercial rather than technical: most of these tools sell purpose-built CRM and outreach connectors as the product, and an open tool-calling protocol makes those connectors easier to route around.
What a mention does not mean
24 of the 94 listings in our index carry no mention of MCP in their own recorded key features or integrations, at collector run 221173 of 27 August 2026: Aider, Artisan, Auto-Respond, Bland AI, Browserbase, Decagon, Deepgram Voice Agent API, GC AI, Julius AI, Junie, Kilo Code, Moveworks, Muse Code, OpenClaw, Qualified, Rox, SkipCalls, Synthflow AI, Unify, Vecbase, Vellum, Warmly, Zendesk AI Agents and Zot.
That is not a count of tools that lack MCP support, and a text match cannot make that distinction. A missing mention means one of several different things: the vendor has not shipped MCP support, the vendor has shipped it and we have not yet written it into the listing’s key features or integrations, or the vendor describes it as planned rather than shipped, as Moveworks does on its own blog, where MCP is framed as a capability it is developing rather than one it has released. If you are evaluating a specific tool on this list, read its current docs directly rather than reading its absence here as a verified No.
AgentsKit shows the limit of this method from the other side. Its own recorded key features contain the phrase “doc bridge mcp server,” so this instrument counts it among the 70 that mention MCP. Its own documentation, though, describes its /api/mcp endpoint as a discovery endpoint that indexes its own catalogue, which our earlier research found to be neither an MCP client nor an MCP server. A word match cannot draw that distinction. Only reading the vendor’s own docs can, which is why this page no longer claims to draw it at the scale of 94 listings.
What the July 2026 pass measured, and why it is not a trend line
We first ran this count on 25 July 2026, against a 45-listing index, using a manual pass that checked every unconfirmed listing against the vendor’s current docs, GitHub organisation or changelog, and recorded 35 of 45 documenting MCP that way. Do not read 35 of 45 against 70 of 94 as a change in adoption, or as two points on one series. The two figures were produced by different methods, one a verified per-listing judgement and the other a text match over recorded fields pinned to a single collector run, and this page does not treat them as comparable.
That July pass also recorded individual verified verdicts, both positive and negative, on ten named tools. None of those verdicts appears on this page: this edition performs no per-listing verification at all, only the single automated text match stated in the Method section below.
How to check MCP support yourself before you build on it
A vendor’s homepage badge is not a working connector. Before you rely on MCP support as a buying decision:
- Look for the vendor’s own MCP server repository, usually on GitHub, or an MCP section in their developer docs. A marketing mention is not either of those.
- Check whether the tool ships as an MCP client, which connects out to other tools’ servers, or an MCP server, which other agents connect into, or both. They are different capabilities and vendors do not reliably distinguish them.
- Test the connection against a real MCP server before you commit an integration. Implementations vary in how complete they are.
Method and sources
This is not a verified per-listing claim. It is a single automated check, run once against a fixed collector run: for each Published listing in The Agents Index at collector run 221173 (27 August 2026), we search the listing’s own recorded attributes, key features and integrations text for the string “MCP,” case-insensitive. A listing counts as mentioning MCP if that text appears anywhere in those fields, whether the mention is a paragraph documenting a shipped server or a single bullet on a pricing page. It does not confirm that the mention is accurate, current, or checked against the vendor’s own docs by a person; an earlier edition of this page did that work by hand and called the result “documenting” MCP support, and that per-listing judgement is not stored anywhere a collector run can pin, so it cannot be reproduced the way every other number on this page can. This edition counts what the recorded text says instead.
At collector run 221173, the attributes field alone matches none of the 94 listings; 7 of the 70 mentions are carried in key features only, 33 in integrations only, and 30 in both.
The figures above were computed from the Published corpus of The Agents Index as it stood on 27 August 2026: 95 published listings, 94 of them with a deployment mode of their own, of which 70 mention MCP and 24 do not. The 95th is Agentix Labs, our one agency listing, held outside this count for the reason stated above. To re-derive any figure above, take the population from the published listings of run 221173, and the attributes, key features and integrations text from the newest listing record at or before run 220483. One narrower reading exists and is not the one this page uses: matching only the integrations field returns 63 at this same run, which happens to be the number this page printed before this correction. That reading disagrees with the withdrawn edition’s own named lists on ten listings, so it is not used here.
Primary sources on MCP itself: Anthropic, “Introducing the Model Context Protocol”, 2024-11-25; Model Context Protocol documentation; OpenAI, MCP in the OpenAI API.
Get the next report
New agents rankings and fresh data reports. One short email, one-click unsubscribe.