langchain-ai · Florence Morris (fjmorris) · Apr 22, 2026 · Apr 22, 2026 · Apr 22, 2026 · Apr 22, 2026
diff --git a/src/docs.json b/src/docs.json
@@ -949,7 +949,8 @@
               "pages": [
                 "langsmith/fleet/quickstart",
                 "langsmith/fleet/essentials",
-                "langsmith/fleet/templates"
+                "langsmith/fleet/templates",
+                "langsmith/fleet/comparison"
               ]
             },
             {

diff --git a/src/langsmith/fleet/comparison.mdx b/src/langsmith/fleet/comparison.mdx
@@ -0,0 +1,182 @@
+---
+title: Comparison with Claude Cowork, Amazon Quick, and Google Workspace Studio
+sidebarTitle: Comparison
+description: Compare LangSmith Fleet with Claude Cowork, Amazon Quick, and Google Workspace Studio to choose the right enterprise agent platform for your team
+---
+
+Use this page to compare [LangSmith Fleet](/langsmith/fleet/index) with [Claude Cowork](https://www.anthropic.com/product/claude-cowork), [Amazon Quick](https://aws.amazon.com/quick/), and [Google Workspace Studio](https://workspace.google.com/studio/).
+
+- **Choose Fleet** if you want to build and share purpose-built agents across your organization, stay model-agnostic, and keep full observability via LangSmith. Fleet is the only option with a self-hosted deployment path and the ability to export agents to code via [Deep Agents](/oss/deepagents/overview).
+- **Choose Claude Cowork** if you want to delegate open-ended tasks to Claude from the desktop for personal knowledge work, and on-device data storage meets your privacy requirements.
+- **Choose Amazon Quick** if you are already on AWS and want an AI assistant with direct access to your AWS data sources and enterprise integrations.
+- **Choose Google Workspace Studio** if your organization runs on Google Workspace and you want no-code agents that work natively inside Gmail, Drive, and Sheets without leaving the Google ecosystem.
+
+## How Fleet differs
+
+### Target users
+
+Fleet covers both org-wide and personal use cases. Teams can build purpose-built agents to share across an organization (for example, a vendor intake agent that serves an entire ops org, or a weekly report agent that saves every account manager thirty minutes on Monday morning), and any user can get help with any task using any tool via Fleet's general-purpose default chat. Cowork is designed for individual workflows: each user runs Claude on their own machine for open-ended task delegation, rather than building and sharing agents across a team. Quick and Workspace Studio are workforce-wide platforms built around answering questions and automating workflows, but not primarily designed for building and sharing purpose-built agents.
+
+### Enterprise controls and access
+
+Fleet provides RBAC, attribute-based access control, and per-agent sharing permissions (Clone, Run, and Edit) that Cowork does not offer. This is a significant difference for organizations that need to manage which teams can access, run, or modify specific agents. Quick and Workspace Studio inherit access controls from their respective cloud platforms (AWS IAM and Google Workspace roles), but neither offers Fleet's per-agent permission model.
+
+Fleet manages spending at the workspace level rather than through per-user or per-team caps. Cowork, Quick, and Workspace Studio all offer more granular spend controls. For enterprise billing options, [contact sales](https://www.langchain.com/contact-sales).
+
+### Model flexibility
+
+Fleet supports any LLM via the OpenAI or Anthropic chat spec, including models from OpenAI, Anthropic, Google, and self-hosted providers. Cowork is locked to Claude and Workspace Studio to Gemini 3. Amazon Quick's supported models are not confirmed from public documentation.
+
+### Ecosystem lock-in
+
+Google Workspace Studio is included in Google Workspace Business and Enterprise plans with no separate purchase, but it works only within the Google ecosystem. Amazon Quick requires existing AWS infrastructure and does not include a bundled productivity suite. Fleet and Cowork are largely ecosystem-agnostic.
+
+### Memory, self-updates, and learning
+
+Fleet agents can persist context across conversations using a dedicated memory system, and can update their own instructions, add tools, or remove tools as they learn from interactions. Of the four platforms compared here, only Fleet supports agent self-modification.
+
+### Observability and governance
+
+Fleet's clearest advantage is its native connection to LangSmith. Every agent run is traced in LangSmith, making it easy to debug performance and run evaluations at scale. The other three have audit and activity logs, but none match Fleet's depth of tracing, evaluations, and debugging through a dedicated observability platform.
+
+### Human-in-the-loop controls
+
+Fleet lets you set tool-level approval requirements so agents check with you before executing sensitive steps, with a [centralized inbox](https://smith.langchain.com/agents/inbox) for reviewing, editing, and approving actions. Cowork takes a similar approach with a "show me the plan first" model. Quick offers notifications but without a centralized inbox for reviewing actions across all agents.
+
+### Code export
+
+Fleet lets you export any agent you build to code via [Deep Agents](/oss/deepagents/overview), the open-source agent runtime that Fleet runs on. Exported agents are MIT-licensed and can be deployed independently of Fleet, modified in code, or integrated directly into your own applications via the [API](/langsmith/fleet/code). None of the other platforms in this comparison offer a code export path.
+
+### Hosting and compliance
+
+Fleet is the only platform in this comparison with a self-hosted deployment option. For teams with compliance requirements, self-hosted and BYOC (bring your own cloud) configurations let you run Fleet entirely within your own infrastructure. Amazon Quick offers enterprise-grade AWS infrastructure natively. Google Workspace Studio adheres to Google Workspace data commitments. Cowork stores data locally on device, making it the only platform in this comparison that keeps data entirely off external servers.
+
+## Overview
+
+- ✅ Available
+- ❌ Not available
+- ⚠️ Partial or limited
+- — Not confirmed from public documentation
+
+<div className="compact-first-col">
+
+| **Aspect** | **LangSmith Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| - | ------------------- | ----------------- | ---------------- | --------------------------- |
+| **Primary use case** | Teams building purpose-built agents to share across an organization, with no-code creation and code export for custom deployments; individuals using a general-purpose chat agent for any task | Individual knowledge work from the desktop (research, analysis, task delegation) | Enterprise AI assistance with deep AWS data and service integration | No-code agents tightly integrated with Google Workspace apps (Gmail, Drive, Sheets) |
+| **Model support** | Model-agnostic: any LLM with an OpenAI-compatible or Anthropic-compatible API | Claude only | — | Gemini 3 |
+| **Interface** | Web app, Slack app, Teams app, API | Desktop app (macOS, Windows) | Web app | Web app (Google Workspace) |
+| **Deployment** | Cloud (LangSmith) or self-hosted | Local on device | Cloud (AWS-hosted) | Cloud (Google-hosted) |
+| **Self-hosting** | ✅ [Beta](/langsmith/deploy-self-hosted-full-platform#enable-fleet-insights-and-polly), [contact sales](https://www.langchain.com/contact-sales) for production readiness details | ❌ | ❌ | ❌ |
+| **Code export** | ✅ [Export to Deep Agents](/langsmith/fleet/code) | ❌ | ❌ | ❌ |
+| **Observability** | LangSmith tracing and evaluations at scale | — | — | Basic run history |
+| **Platform license** | Proprietary | Proprietary | Proprietary | Proprietary |
+| **Code export license** | MIT ([Deep Agents](/oss/deepagents/overview)) | N/A | N/A | N/A |
+
+</div>
+
+## Compare features
+
+- ✅ Available
+- ❌ Not available
+- ⚠️ Partial or limited
+- — Not confirmed from public documentation
+
+### Agent capabilities
+
+<div className="compact-first-col">
+
+| Feature | **Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| ------- | --------- | ----------------- | ---------------- | --------------------------- |
+| General-purpose chat agent | ✅ [Fleet chat](https://smith.langchain.com/agents) | ✅ | ✅ | ✅ |
+| No-code agent builder | ✅ | ❌ | ❌ | ✅ |
+| Slack-native integration | ✅ [Native Slack app](/langsmith/fleet/slack-app) | ❌ | ❌ | ❌ |
+| Microsoft Teams integration | ✅ [Teams app](/langsmith/fleet/teams-app) | ❌ | ❌ | ❌ |
+| Scheduled runs | ✅ [Schedules](/langsmith/fleet/schedules) | ✅ | ❌ | ✅ |
+| Sub-agents | ✅ [Sub-agents](/langsmith/fleet/essentials#sub-agents) | ✅ Managed Agents | — | ❌ |
+| Skills system | ✅ [Skills](/langsmith/fleet/skills) | ✅ | ❌ | ❌ |
+| Human-in-the-loop | ✅ [Central approvals inbox](/langsmith/fleet/essentials#human-in-the-loop) | ✅ Plan-first approach | ⚠️ Notifications only | — |
+| MCP client | ✅ [Remote MCP servers](/langsmith/fleet/remote-mcp-servers) | ✅ | ✅ | ❌ |
+| Web search | ✅ (via Exa, Tavily) | ✅ Built-in | ✅ Built-in | — |
+
+</div>
+
+### Memory and self-updates
+
+<div className="compact-first-col">
+
+| Feature | **Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| ------- | --------- | ----------------- | ---------------- | --------------------------- |
+| Long-term memory | ✅ [Persistent memory files across sessions](/langsmith/fleet/essentials#memory) | ✅ | — | ❌ |
+| Thread-scoped context | ✅ | ✅ | ✅ | ✅ |
+| Self-updating agents | ✅ [Agents can add tools, remove tools, and update their own instructions](/langsmith/fleet/essentials#self-updates) | ❌ | ❌ | ❌ |
+| Approval gate for memory writes | ✅ [Configurable per agent](/langsmith/fleet/manage-agent-settings) | ❌ | ❌ | ❌ |
+
+</div>
+
+### Enterprise controls
+
+<div className="compact-first-col">
+
+| Feature | **Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| ------- | --------- | ----------------- | ---------------- | --------------------------- |
+| Role-based access control | ✅ [RBAC with per-tool permissions](/langsmith/rbac) | ✅ SCIM-based groups and roles | ✅ AWS IAM | ✅ Google Workspace roles |
+| Attribute-based access control | ✅ [Per MCP server and integration](/langsmith/fleet/access-and-oversight#attribute-based-access-control) | ❌ | ❌ | ❌ |
+| Per-agent sharing and permissions | ✅ [Clone, Run, and Edit access per agent](/langsmith/fleet/access-and-oversight#permissions-and-sharing) | ❌ | ❌ | ❌ |
+| Credential model (fixed or per-user) | ✅ [Configurable per agent](/langsmith/fleet/access-and-oversight#agent-identity-and-credentials) | ✅ | ✅ | ✅ |
+| Spend limits | ⚠️ Managed at workspace level | ✅ | ✅ | ✅ |
+| SCIM provisioning | ✅ | ❌ | — | — |
+| Audit trail | ✅ [Structured LangSmith traces](/langsmith/fleet/access-and-oversight#observability-and-audit-trail) | ❌ | — | ⚠️ Basic run history |
+
+</div>
+
+### Deployment and hosting
+
+<div className="compact-first-col">
+
+| Feature | **Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| ------- | --------- | ----------------- | ---------------- | --------------------------- |
+| Cloud-hosted | ✅ | ❌ | ✅ | ✅ |
+| Self-hosted | ✅ [Beta](/langsmith/deploy-self-hosted-full-platform#enable-fleet-insights-and-polly), [contact sales](https://www.langchain.com/contact-sales) for production readiness details | ❌ | ❌ | ❌ |
+| Custom models | ✅ [Any OpenAI- or Anthropic-compatible API](/langsmith/fleet/essentials#custom-models) | ❌ | ⚠️ Custom models via Bedrock | ❌ |
+| Call agents from your app | ✅ [API access](/langsmith/fleet/code) | ✅ Managed Agents API | ✅ Amazon Quick API | ❌ |
+| Export to code | ✅ [Export to Deep Agents](/langsmith/fleet/code) | ❌ | ❌ | ❌ |
+
+</div>
+
+### Integrations and tools
+
+<div className="compact-first-col">
+
+| Feature | **Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| ------- | --------- | ----------------- | ---------------- | --------------------------- |
+| Google Workspace (Gmail, Drive, Sheets, Docs) | ✅ | ✅ Plugin | ✅ | ✅ Native |
+| Microsoft 365 (Outlook, Teams, SharePoint, Excel) | ✅ | ✅ Plugin | ✅ | ❌ |
+| GitHub | ✅ | ✅ Plugin | ✅ | — |
+| Slack | ✅ Native | ❌ | ❌ | ❌ |
+| CRM (Salesforce, HubSpot) | ✅ | ✅ Plugin | ✅ | ✅ |
+| Project management (Linear, Jira, Notion) | ✅ | ✅ Plugin | ✅ | ✅ |
+| Custom tools via MCP | ✅ | ✅ | ✅ | ❌ |
+| Webhooks | ✅ [Webhooks](/langsmith/fleet/webhooks) | ❌ | ❌ | — |
+
+</div>
+
+<Tip>
+  In the integrations table above, ✅ indicates the integration is available. Depth and supported actions vary by platform. See [Fleet tool integrations](/langsmith/fleet/tools) for the full list of Fleet's built-in integrations and what each one can do.
+</Tip>
+
+### Observability
+
+<div className="compact-first-col">
+
+| Feature | **Fleet** | **Claude Cowork** | **Amazon Quick** | **Google Workspace Studio** |
+| ------- | --------- | ----------------- | ---------------- | --------------------------- |
+| Native tracing | ✅ [LangSmith traces for every run](/langsmith/observability) | — | — | ⚠️ Basic |
+| Evaluations | ✅ [LangSmith evaluations](/langsmith/evaluation-concepts) | ❌ | ❌ | ❌ |
+
+</div>
+
+For pricing and SLA information, [contact sales](https://www.langchain.com/contact-sales).
+
+<Note>
+    Last updated April 21st, 2026. These products evolve quickly. If something has changed, please [file an issue](https://github.com/langchain-ai/docs/issues) to help us keep this page current.
+</Note>