AI Agent Tools With MCP Support
49 of the 64 tools in our index document real support for the Model Context Protocol (MCP) today — as a client, a server, or both. Coding agents and frameworks lead; sales and marketing agents are the clear laggard.
MCP (the Model Context Protocol) is becoming the default way an AI agent reaches your tools and data instead of a bespoke integration per pair — see our full explainer for what it is and why it matters when you're picking an agent to build on. This collection answers the practical question that piece leaves open: which tools in our own index actually document it today. 49 of the 64 tools here do, spanning every one of our seven categories — some as a client (they can call out to any MCP server), some as a server (other agents can call into them), several as both — including Glean, which ships MCP in both directions today.
The split isn't even. Agent frameworks are unanimous (11 of 11 — Google ADK and Claude Agent SDK, both major-lab arrivals since the last count, keep the category perfect) and coding agents are close behind (16 of 19) — unsurprising, since MCP began as a developer-tooling protocol. Sales and marketing agents are the clear laggard: only Clay (a full MCP server) and 11x (one undetailed pricing-page bullet, the thinnest evidence in this set) document anything at all. Fourteen tools in the index are not included here — Aider, Lindy, GC AI, Artisan, Qualified, Rox, Decagon, Bland AI, Juggler, Zot, Unify, Moveworks, Undermind and Vecbase — either because they have confirmed they don't support it yet (Aider has an open, unresolved GitHub request since December 2024; Lindy's own blog says it isn't compatible yet; Moveworks' own blog still frames MCP as "in development," not shipped) or because no MCP claim appears anywhere in their own docs, site or changelog as of this check.
Why this collection
A tool earns a place here only when its own docs, GitHub repo, blog or changelog make an explicit MCP claim — as a client, a server, or both — not a generic "integrations" or "API" mention. 35 of the 49 here were confirmed in our original MCP census (see the full article for the per-tool sourcing); the other 14 — Warp, CodeRabbit, Salesforce Agentforce, Antigravity CLI, PolyAI, Mastra, Letta, Factory, Glean, Zed, Microsoft Agent Framework, Botsify, Google Agent Development Kit and Claude Agent SDK — were added to the index afterward. The first 9 of those 14 were independently re-verified against primary vendor docs on 2026-07-25, the same pass that checked Unify and Moveworks and confirmed neither supports MCP yet; Zed, Microsoft Agent Framework, Botsify, Google Agent Development Kit and Claude Agent SDK were each confirmed directly from their own already-Published, sourced listing fields (integrations and key-features), not a fresh vendor-docs pass — each already names MCP explicitly as a documented integration, not a generic API mention. Every one of these facts is also recorded on the tool's own listing (integrations + sourced citation), not stated here alone. MCP adoption is moving fast in 2026; treat this as a snapshot, not a ceiling, and check a specific vendor's current docs before relying on it for a real integration.
- Cursor
The AI code editor, an agent that reads, writes and edits across your whole codebase.
MCP servers and tools are a core integration surface built into the editor.
- Claude Code
Anthropic’s terminal-native coding agent with full-repository context.
MCP servers and external tools are a core, first-class way to extend what it can reach.
- Cline
The open-source coding agent, an autonomous, bring-your-own-key agent that lives in your editor, terminal or SDK.
First-class support plus its own in-app MCP marketplace for one-click server installs.
- Windsurf (now Devin Desktop)
The agentic IDE that was Codeium, then Windsurf, now Cognition’s Devin Desktop, a command centre for coding agents.
MCP servers are a core integration, carried over unchanged into the Devin Desktop rebrand.
- OpenCode
A free, open-source (MIT) AI coding agent for the terminal, desktop and IDE, bring any of 75+ model providers, including your own local Ollama models.
A dedicated MCP-server integration layer sits alongside its own custom agent-skills system.
- Bolt.new
StackBlitz's prompt-to-app builder that runs a real Node.js environment inside the browser tab itself, no remote VM.
Connects out to MCP servers (Notion, Linear, GitHub, custom) as a client — no server of its own.
- OpenAI Codex CLI
OpenAI's open-source (Apache-2.0) terminal coding agent, bundled into ChatGPT plans from Free through Pro — Business seats are now grandfathered-only.
MCP servers and tools are documented alongside its sandboxed execution model.
- Devin
Cognition’s autonomous AI software engineer, delegate a ticket, get a reviewed pull request back.
An MCP Marketplace ships as a client, connecting Devin to external MCP servers.
- GitHub Copilot
The mature AI pair-programmer, now with an agent mode, deep in the GitHub + IDE stack.
Both Copilot Chat and the autonomous coding/cloud agent connect out to external MCP servers.
- OpenHands
Open source AI agent platform for cloud coding agents, from free local use to enterprise self-hosted deployment.
An MCP client over SSE, Streamable HTTP or Stdio transports, configured directly.
- Replit Agent
Describe an app and Replit Agent builds, tests and deploys it in your browser, a full-stack app factory for anyone.
Documented as an MCP client in Replit's own developer docs.
- Factory
Agent-native software development — specialized "Droid" agents that triage, code, test, review, document and ship changes across the SDLC, from a $20/mo CLI to a fully managed enterprise cloud.
Droid is an MCP client with a built-in registry of 40+ pre-configured servers, browsable via `/mcp`.
- CodeRabbit
An AI code-review agent, not a code-writing one — reviews every pull request automatically across GitHub, GitLab, Bitbucket and Azure DevOps, with a genuine free tier for unlimited repos.
Connects out to MCP servers as a client only — plan tier caps how many, from 5 on Pro up to 20 on Enterprise.
- Antigravity CLI
Google's terminal-native coding agent that replaced the open-source Gemini CLI in June 2026 — same Antigravity 2.0 agent core, now closed-source and tied to a Google AI subscription.
Ships its own local MCP server exposing tools and prompts, plus first-party servers for Google Workspace and Data Cloud.
- Warp
The AI-native terminal that replaces your shell and orchestrates Claude Code, Codex and its own Warp Agent side by side.
Its Oz agent CLI takes MCP server config directly — by UUID, inline JSON or file path — to reach tools like GitHub, Linear and Sentry.
- LangChain
The most widely-used framework for building LLM apps and agents, with LangGraph for stateful control.
The official `langchain-ai/langchain-mcp-adapters` package makes LangChain agents a fully-supported MCP client.
- CrewAI
A popular open-source Python framework for role-based, collaborating multi-agent teams, free to self-host, now a stable 1.0 release with an optional managed AMP cloud layer.
An official `MCPServerAdapter` connects crews to external MCP servers as a client.
- LlamaIndex
The data framework for LLM apps, RAG-first, now with agent workflows and a document-parsing cloud.
Both directions — consumes external MCP servers and can publish its own workflows as one.
- OpenAI Agents SDK
OpenAI’s lightweight open-source framework for building agents, handoffs, guardrails and tracing in a few primitives.
MCP servers are a first-class external tool source, alongside human-in-the-loop approval.
- Pydantic AI
The type-safe agent framework from the Pydantic team, that FastAPI feeling for building AI agents.
MCP servers are a first-class tool source for its agents.
- Microsoft AutoGen
Microsoft’s open-source framework for orchestrating conversations between multiple AI agents.
The official `autogen-ext` package is an MCP client over STDIO, SSE or Streamable HTTP.
- Mastra
The TypeScript-native agent framework — agents, durable workflows, memory and observability in one Apache-2.0 package, with an optional managed cloud platform.
Both directions via dedicated packages — `MCPClient` connects out, `MCPServer` exposes Mastra's own tools to any MCP client.
- Letta
A stateful-agent framework — open-source, memory-first agents whose context, tools and identity persist as a durable server object instead of resetting between sessions.
Acts as an MCP host, connecting agents to external servers' tools, resources and prompts.
- Dify
Open-source, visual platform for building production LLM apps and agents, self-host it or run it in the cloud.
Any Dify app can be published as an MCP-compatible tool.
- Relevance AI
Build and run an "AI workforce", teams of no-code agents that do real sales, ops and research work.
Agent actions inside the platform are directly extendable by connecting them to external MCP servers.
- n8n
Source-available workflow automation with native AI-agent nodes, self-hostable and integration-rich.
Both directions — a dedicated MCP Server Trigger node and an MCP Client node.
- Salesforce Agentforce
Salesforce's enterprise agentic-AI platform (rebranded "Agentforce 360") — prebuilt Service, SDR, Sales Coach and Commerce agents plus a low-code builder, running on live CRM data.
A native MCP client connects to any MCP-compliant server, plus verified AgentExchange partner servers.
- Perplexity
The AI answer engine, with agentic Deep Research and the Comet browser that acts on the web for you.
Both directions — an official MCP server, plus local MCP client support.
- Manus
An autonomous general AI agent that plans and executes multi-step tasks end-to-end in its own cloud computer.
Prebuilt MCP connectors (client) to Notion, Linear, Gmail, Stripe and more.
- GPT Researcher
An open-source autonomous research agent that plans, searches many sources and writes a cited report.
An MCP server is available via a companion project (`gptr-mcp`), not bundled in the main repo.
- Elicit
An AI research agent built for scientific literature, search, screen and extract cited data from 138M+ papers.
An official MCP server exposes its own search and extraction API.
- Vapi
The developer platform for voice AI agents, bring your own model, voice and telephony, and Vapi orchestrates the call.
Custom tools are reachable through both native function calling and Model Context Protocol servers.
- ElevenLabs Conversational AI
The voice-quality specialist’s own agent layer, 10,000+ voices and instant cloning, with a bring-your-own-LLM brain.
Both directions — connects to external MCP servers and ships its own official server.
- Retell AI
A voice AI platform for human-sounding phone agents, build a call flow, plug in your model and telephony, ship.
Both directions — an MCP server for managing agents, plus mid-call client tool calls.
- LiveKit Agents
The open-source, self-hostable voice agent framework, Apache 2.0, provider-agnostic, and the infrastructure behind ChatGPT’s Advanced Voice Mode.
First-class client support, documented for both HTTP and stdio transports.
- PolyAI
The pre-LLM-era enterprise voice AI incumbent — $750M-valued, 200+ enterprise customers — that opened a genuine self-serve, free-to-start tier in May 2026.
Runs its own MCP server abstracting storage and data access, on top of broader MCP-style integrations across the platform.
- Sierra
Bret Taylor’s enterprise agent platform, branded AI agents that resolve customer issues end to end, priced on outcomes.
Agents are natively publishable to ChatGPT Apps via MCP (server direction).
- Intercom Fin
Intercom’s customer-service AI agent, resolves tickets omnichannel over any helpdesk, billed per resolution.
A fully documented API, data connectors and MCP support for deeper integration.
- Crescendo
Fully-managed AI-native customer service, replaces your entire CX stack with agents, QA and human-in-the-loop support, billed per resolution.
Connects its Assistant to external MCP servers, and exposes its own MCP config service.
- Cresta
Unifies an autonomous AI Agent, real-time human Agent Assist and Conversation Intelligence on one contact-center platform, enterprise, custom-quoted, no self-serve tier.
MCP for standardized tool access, alongside its existing API-based CRM/billing integrations.
- Ada
A Toronto-founded, $1.2B-valued AI customer-service agent platform predating the LLM wave — sales-led, no public pricing, a genuine public API.
Ships its own MCP server, letting external AI assistants (Claude Desktop, ChatGPT) query its systems directly.
- Clay
GTM data platform with an AI research agent (Claygent), enrich, research and personalise outbound at scale.
Runs its own MCP server so Claude, ChatGPT, Copilot and Glean can query a Clay workspace directly.
- 11x
AI "digital workers" for sales, Alice runs autonomous outbound; Julian is an AI voice agent.
'Custom MCP Integration' is listed on the Enterprise pricing tier — the thinnest evidence here: no docs, no stated client/server direction.
- Glean
Enterprise "Work AI" platform — permission-aware search across 100+ systems, an AI assistant and a no-code agent builder, all grounded in one Knowledge Graph. $7.2B valuation, June 2025.
Ships MCP in both directions today — an MCP server exposing Glean's own search, chat and knowledge graph to external tools (Cursor, VS Code, Claude Desktop, ChatGPT), and an MCP host connecting 17+ preloaded remote servers.
- Zed
The open-source, GPU-native code editor built from scratch by Atom's creators — runs Claude Code, Codex and other agents via the open Agent Client Protocol it created. Free forever; Pro from $10/month.
MCP servers extend the agent's available context and tools — a client integration listed alongside its ACP support for external coding agents like Claude Code and Copilot.
- Microsoft Agent Framework
Microsoft’s production-ready, open-source successor to AutoGen and Semantic Kernel — one agent SDK for .NET, Python and Go.
Native Model Context Protocol tool support ships as a named feature alongside Agent2Agent (A2A) protocol support — one of the framework's two headline 2026 interoperability standards.
- Botsify
A decade-old, bootstrapped no-code AI agent builder for support, lead-gen and appointment booking across chat channels — with a distinct white-label track for agencies to resell under their own brand.
MCP, document and web-search grounding ships as a named feature on both the Done-for-You and White Label plans, letting the agent reach tools and data beyond a static script.
- Google Agent Development Kit
Google's open-source, code-first agent SDK — four language runtimes, native A2A + MCP protocol support, and an optional managed Agent Runtime on Google Cloud.
Both directions — native Model Context Protocol tool support ships alongside Agent2Agent (A2A) protocol support, including a bundled MCP Toolbox for 30+ databases on the Go SDK.
- Claude Agent SDK
Anthropic's own agent-building library — the same agent loop, tools and context management that power Claude Code, for Python and TypeScript. Claude-only, with no multi-provider escape hatch.
Both directions — external MCP servers connect the same way Claude Code does, and the `@tool` decorator mounts custom functions as an in-process MCP server with no separate subprocess to manage.