Arcade.dev’s agent-optimized MCP tools are now available in LangSmith Fleet. Fleet enables everyone on your team to create, use, and share autonomous agents using natural language. Simply describe what you want. Fleet makes a plan and takes action, working across the apps your team uses daily.
As the MCP runtime, Arcade delivers secure agent authorization, 7,500+ reliable tools, and governance for every action an agent takes. Fleet now includes native Arcade integrations, giving your agents access to the tools and auth infrastructure they need out of the box. To get teams started fast, Arcade provides over 60 production-ready templates for sales, marketing, engineering, and support use cases.
Why This Matters
LangSmith Fleet makes it easy for anyone to build agents, no prompt engineering expertise required. But agents are only as useful as the tools they can access. Most MCP tools out there are just thin wrappers around APIs, built for structured application requests, not the messy natural language that agents produce. The result: inconsistent behavior, cryptic errors, and agents that work in demos but fail in real workflows.
Arcade has built 7,500+ agent-optimized tools across 81 MCP toolkits for the systems teams actually use: GitHub, Slack, Salesforce, Gmail, Linear, and many others. These aren’t API wrappers. When an agent needs to “make the intro paragraph friendlier,” Arcade translates that to the exact segment, index, and text the downstream API requires. The agent never has to reason about endpoint structure, parameter formats, or pagination. Every tool follows consistent patterns with error handling agents understand, so your agents can take complex actions across any system with the accuracy and reliability production workflows demand.
Arcade’s runtime also handles the auth that makes multi-user agents possible in production. Each action enforces least privilege, inheriting the permissions of the specific user the agent is acting for. Every tool call is logged with a full audit trail, giving teams visibility into what each agent did, on whose behalf, against which system.
60+ Templates Ready to Work for You
Skip the blank canvas. Start with battle-tested templates built for real work:
Marketing: Deploy a Social Media Content Autopilot that generates and schedules posts from your content calendar automatically. Launch an AI Newsletter Curator that scans RSS feeds and drafts your weekly roundup while you sleep.
Sales: Spin up a Deal Research Assistant that creates comprehensive prospect dossiers from LinkedIn and CRM data in seconds. Or build a Lead Capture & Routing System that enriches every form submission and routes to the right rep instantly.
Support: Launch a Customer Health Monitor that tracks usage patterns and flags at-risk accounts before they churn. Deploy an NPS Response Handler that routes promoters to review sites and triggers CSM follow-up for detractors automatically.
Engineering: Set up a Weekly Engineering Digest that compiles GitHub and Linear status reports into a polished summary. Build an Incident Response Coordinator that creates war rooms, tracks timelines, and prepares postmortem templates the moment something breaks.
Each template is pre-configured with the right tools and permissions. Need to extend or modify your agent? Just give it feedback like you would a teammate, Fleet uses long-term memory to improve over time.
Get Started Today
Each template connects through an Arcade MCP Gateway - a single endpoint that handles tool access, per-user authorization, and audit logging for every action your agent takes. We’re excited to work with LangChain to put sophisticated agents into the hands of more users and handle real work across the systems that matter.
Sign up for Arcade and LangSmith Fleet for free to start kickstarting your productivity.
FAQ
What is the integration between Arcade and LangSmith Fleet?
Arcade integrates directly with LangSmith Fleet to provide AI agents with over 7,500 Model Context Protocol (MCP) tools and a governed authorization infrastructure. Users build and deploy autonomous agents that reliably take action across external applications such as GitHub, Slack, and Salesforce.
How do Arcade MCP tools differ from standard API wrappers?
Arcade MCP tools are optimized for the unstructured natural language AI agents produce, not rigid application requests. When an agent requests a complex action, Arcade translates the instruction into the exact endpoint structure required by the downstream API. This prevents cryptic errors and unpredictable behavior that occur when agents use basic API wrappers.
How does Arcade handle user authentication for AI agents?
Arcade is the MCP runtime that enforces least-privilege authentication for every action a LangSmith Fleet agent takes. The runtime inherits the permissions of the human user the agent is acting on behalf of and maps them to existing IDP and SaaS policies. Arcade logs every tool call to provide a complete audit trail of what the agent did and which systems it accessed.
What is an Arcade MCP Gateway?
The Arcade MCP Gateway is the access point for AI agents. The Arcade runtime enforces the policy. LangSmith Fleet templates connect through the Gateway so agents interact with production systems under governance. It provides the audit and access controls required for deploying complex AI agents in enterprise environments.
What types of AI agent templates does LangSmith Fleet offer?
LangSmith Fleet offers over 60 production-ready templates powered by Arcade tools for sales, marketing, engineering, and support workflows. Popular templates include social media content autopilots, deal research assistants, customer health monitors, and incident response coordinators. Each template is pre-configured with the exact tools and permissions a team needs to start working.
Can I build AI agents in LangSmith Fleet without prompt engineering skills?
Yes. You build AI agents in LangSmith Fleet by describing your goals in conversational natural language. Tell the runtime what you want, and Fleet plans the actions across your connected applications. The system uses long-term memory to learn from text feedback and improve the agent over time.
