One platform for OpenAI, Anthropic and more

Wide Moat combines multiple providers and supported tools in one platform, adding capabilities beyond standard AI subscriptions.

  • Out of the box
  • Needs extra setup
  • Not available / High risk

Swipe the table sideways to see all three columns.

CapabilityChatGPT BusinessClaude TeamWide Moat
Build
Skill&Agent Factory
Needs extra setup: Plugins on OpenAI Cloud
OpenAI plugins

Package reusable skills and connections to company systems in a plugin, then share it with colleagues or publish it to the workspace directory. ChatGPT uses those instructions and tools to perform tasks. Configure schedules or supported event triggers separately; available triggers depend on the plan and workspace settings. The shared catalog is hosted by OpenAI. Execution depends on the client: cloud tasks run on OpenAI Cloud, while local clients can execute code and tools locally. Running a workflow engine inside your infrastructure requires separate setup.

Needs extra setup: Plugins on Anthropic Cloud
Anthropic plugins

Bundle skills, specialized agent configurations and tool connections in a plugin. Publish it through the organization library; admins can make it available, install it by default or require it. Employees can also publish standalone skills directly, subject to organization policy. The shared catalog is hosted by Anthropic. Execution depends on the client: Claude Code can run plugin tools and subagents locally. A separate workflow engine or customer-hosted execution environment requires additional setup.

Out of the box: Create, share & run inside the platform
Built-in agent factory

Create skills and agents, publish them to a shared registry and give colleagues access. Automate processes with the built-in workflow-agent engine. Agents run in isolated sandboxes inside the platform. The registry, workflow engine and sandboxes are part of the deployed platform. External model calls follow your selected providers and platform rules.

Live
AI BI & Data Factory
Needs extra setup: Data plugins
OpenAI Cloud
OpenAI data analysis

Use the Data plugin with approved database connections to query company data, explore schemas and build charts, dashboards and reports. Connected sources can include BigQuery, Snowflake and Databricks. Source-system permissions and the connection identity determine which data is available; configure shared credentials with care. Connect your sources and supply trusted metric definitions and relationships. dbt is optional, and reading repository code is not a prerequisite. Hosted analysis runs on OpenAI Cloud; local Codex analysis is another execution option. A customer-hosted BI and data-access platform requires separate integration.

Needs extra setup: Data plugins
Anthropic Cloud
Claude data analysis

Use the Data plugin to query connected warehouses, write SQL and build visualizations and dashboards. Source-system permissions and the connection identity determine which data is available; configure shared credentials with care. Connect your sources and supply trusted metric definitions and relationships. dbt is optional, and reading repository code is not a prerequisite. Hosted analysis runs on Anthropic Cloud; local Cowork and Claude Code offer other execution options. A customer-hosted BI and data-access platform requires separate integration. Claude also offers a separate Dashboards feature in beta.

Out of the box: Data access, analytics & BI
Inside the platform
Built-in data tools

Connect company data and manage who can access each connected dataset through the platform and source-system permissions. Use built-in analysis and visualization tools to query data, create dashboards and investigate business questions inside the deployed platform. Company-specific datasets, access policies, metric definitions and calculations still need configuration and validation. Selected external models can receive query results and context through the configured gateway.

Live
Internal apps
UI, sign-in, integrations, data, workflows & controls
Needs extra setup: Sites: light apps on OpenAI Cloud
Sites

Build and host web apps on OpenAI Cloud, including compatible server-side code, a database and sign-in. The hosting runtime has limits. Custom background services or deployment inside your company infrastructure need separate setup. Scheduled ChatGPT tasks are a separate component and need integration with the app.

Needs extra setup: Artifacts: light apps on Anthropic Cloud
Artifacts

Build interactive apps on Anthropic Cloud with connected tools and persistent text storage, limited to 20 MB per artifact. A custom server backend, background services or deployment inside your company infrastructure need separate setup. Scheduled Claude tasks are separate from the artifact.

Out of the box: Build, test, deploy & run inside the platform
Self-service app platform

Build, test and launch internal apps through a self-service workflow. Agents help create the app, review code and run security checks. The platform automates deployment and hosting for the app, its database and background services. Built-in monitoring and scaling help keep it running. Adjust resources and limits through the platform without setting up or maintaining the underlying infrastructure yourself.

Live
Software Factory
Needs extra setup: Coding agents
Locally or OpenAI Cloud
Codex development tools

Develop with Codex locally or in configurable OpenAI Cloud environments. Agents can write, test, debug and review code. Connecting company-managed workspaces, CI/CD, private test environments and release approvals requires setup. The Codex GitHub Action can execute in your own CI with separate API authentication.

Needs extra setup: Coding agents
Locally or Anthropic Cloud
Claude Code development tools

Develop with Claude Code locally or in configurable Anthropic Cloud environments. Agents can write, test, debug and review code, including fixes for failed CI checks. Connecting company-managed workspaces, CI/CD, private test environments and release approvals requires setup. Remote SSH sessions are another option; self-hosted cloud sessions are a Team/Enterprise beta whose runners your team deploys and operates.

Out of the box: AI-powered SDLC inside the platform
AI-powered software lifecycle

Connect repositories, coding agents and CI/CD to develop, test, review, deliver and maintain software. Developers and agents use preconfigured cloud workspaces, isolated sandboxes and test environments inside the platform deployed in your infrastructure. Manage access, resources, security checks and release approvals through the platform. Configure project-specific tests and integrations. Selected external models can receive code and context through the configured gateway.

Live
Secure
Credentials, secrets & personal data
Not available / High risk: No automatic masking
No built-in masking

Sensitive data included in a prompt is sent to OpenAI for processing without a built-in company masking layer in ChatGPT Business. Enterprise partner DLP integrations are separate.

Not available / High risk: No automatic masking
No built-in masking

Sensitive data included in a prompt is sent to Anthropic for processing without a built-in company masking layer in Claude Team. Claude Code file-access rules can restrict access, but do not mask the contents of allowed requests.

Out of the box: Automatic masking
Masking before sending

Wide Moat masks credentials, secrets and personal data before sending requests to the model. The configured rules determine which data is masked, and apply to requests sent through the platform.

Live
Where AI processes files, code and sensitive data
Not available / High risk: Locally or OpenAI Cloud
Local and cloud execution

Local tools such as Codex work on employees’ computers; hosted tasks run in OpenAI Cloud. Keeping more sensitive files, credentials and local services on a laptop can increase exposure if it is lost or compromised. Depending on granted access, faulty code can damage files or disrupt the local working environment. Device security, encryption, backups and sandboxing affect the risk. File content and tool results included in model requests can also reach OpenAI.

Not available / High risk: Locally or Anthropic Cloud
Local and cloud execution

Local Claude Code tools work on employees’ computers; hosted tasks run in Anthropic Cloud. Keeping more sensitive files, credentials and local services on a laptop can increase exposure if it is lost or compromised. Depending on granted access, faulty code can damage files or disrupt the local working environment. Device security, encryption, backups and sandboxing affect the risk. File content and tool results included in model requests can also reach Anthropic. Customer-hosted Code runners require a separate beta setup.

Out of the box: Inside the platform
Isolated work environments

Files, code and agent tools run in dedicated work environments inside your platform, rather than on an employee’s main computer. With ephemeral environments configured, the platform removes them after use and can create a fresh one if an environment breaks. Keep required files and results in persistent storage and backups; recreating an environment does not recover unsaved work. Isolation limits damage within the assigned permissions, but connected systems still need access controls. External models can receive request content after configured masking rules apply.

Live
Access to internal systems
Needs extra setup: Internet access or secure tunnel
Central connection

An administrator adds the connector once for the workspace. Employees use the shared connection and sign in when required; they do not install an MCP server on each computer. The connector server must be reachable from OpenAI Cloud, directly or through Secure MCP Tunnel. The tunnel needs a separate setup and keeps the server private without opening inbound ports. Restrict permissions, protect credentials and monitor access; misconfiguration can expose internal data.

Needs extra setup: Internet access required
Central connection

An owner adds the remote connector once for the organization. Employees enable it and sign in when required; they do not install an MCP server on each computer. The connector server must be reachable over the internet from Anthropic Cloud. This does not mean exposing the entire internal system or allowing anonymous access. Protect the connection with authentication, restricted permissions and firewall rules; misconfiguration can expose internal data.

Out of the box: Inside the platform
Central connection inside the platform

Configure a shared connector inside the customer-deployed platform and grant employees access. The connector reaches internal databases and tools from there, without exposing those systems to the internet for this connection. Restrict permissions and protect credentials. External models can still receive data included in requests, subject to configured masking rules.

Live
Custom guardrails
Needs extra setup: Requires AI Gateway setup
Custom gateway setup

Route supported Codex clients through a gateway your organization deploys and operates. Implement custom request checks, blocking rules and alerts on that server, then distribute the connection settings to employees. This requires gateway hosting, configuration and integration. A Codex gateway does not automatically cover ordinary ChatGPT web chats or every OpenAI product.

Needs extra setup: Requires AI Gateway setup
Custom gateway setup

Route Claude Code through a gateway your organization deploys and operates. Implement custom request checks, blocking rules and alerts on that server, then distribute the connection settings to employees. This requires gateway hosting, configuration and integration. Claude Code gateway settings do not automatically cover ordinary Claude web chats or every Claude product. Organization-wide Inference hooks are a separate Enterprise capability, outside the Team plan.

Out of the box: Custom checks & alerts
Configured in the platform
Built-in gateway controls

Configure custom checks and alerts through the platform AI gateway. Apply company rules to requests, responses and supported actions, including blocking or processing data and notifying the responsible team when a rule matches. Checks apply to traffic routed through the gateway; rule logic and notification integrations need configuration. The gateway runs inside the deployed platform.

Live
Control & Manage
Model choice
Not available / High risk: OpenAI only
OpenAI only

ChatGPT offers OpenAI models.

Not available / High risk: Anthropic only
Anthropic only

Claude offers Anthropic models.

Out of the box: Any model
(API · Subscription · Local)
Model and provider choice

Connect supported models through provider APIs, subscription-based access or OpenRouter. Run open models on your own hardware or rented GPUs, including servers using vLLM or SGLang. Available models and features depend on provider access, subscription terms and your deployment configuration.

Live
AI cost optimization
Not available / High risk: Manual model selection
Spend limits
OpenAI cost controls

Choose models and set monthly limits on additional workspace credit usage, including per-user limits. Native automatic mode switching follows OpenAI behavior; the Business controls do not provide an administrator-configurable gateway that routes company requests among models by cost rules. Red refers to that automatic cost-routing capability.

Not available / High risk: Manual model selection
Spend limits
Anthropic cost controls

Choose models and set monthly extra-usage spend limits for the organization and individual members. Claude Code also offers opusplan, which uses Opus for planning and Sonnet for execution. This specific mode does not provide a central gateway that routes company requests among models by configurable cost rules. Red refers to that automatic cost-routing capability.

Out of the box: Spend limits &
Automatic model routing
Central cost optimization

Set spend limits and configure automatic model routing at the platform gateway. Route requests among connected providers according to configured rules to balance cost and task needs. Controls cover traffic routed through the gateway. Savings depend on model prices, workload and routing configuration.

Live
Compute resource management
Needs extra setup: Limited
Provider-managed resources
Compute resource controls

Codex Cloud provides isolated task VMs with provider-defined resources, and local Codex uses the machine on which it runs. Configuring a shared customer-operated pool with workspace sizes, capacity and resource quotas requires separate infrastructure and integration. Custom CI runners and remote hosts are separate setups; this differs from managing a company-wide compute platform in the Business subscription.

Needs extra setup: Limited
Provider-managed resources
Compute resource controls

Anthropic manages the resources of its hosted Claude Code sessions; local and SSH sessions use their host resources. Team and Enterprise offer self-hosted cloud sessions in public beta, where your team deploys runners and manages their capacity and scaling. This is additional infrastructure to operate, rather than a ready company-wide resource platform included with Team.

Out of the box: Capacity, quotas & limits
Configured in the platform
Platform resource management

Configure the resources for cloud workspaces, agent sandboxes and application execution inside the platform. Set compute and memory allocations, storage quotas and concurrency limits, and plan capacity for expected demand. Scale or recreate environments within the infrastructure available to your deployment. Resource management is separate from model-token budgets; available capacity and resilience depend on the deployment configuration.

Live
Custom controls & usage analytics
Needs extra setup: Telemetry & reporting setup
Custom usage monitoring

Standard usage and credit reports are available. For custom monitoring of supported Codex sessions, configure telemetry export and connect a collector, storage and reporting tools. Build the dashboards, alert rules and notification integrations you need. Available events and prompt content depend on export settings and permissions. Codex telemetry does not cover ordinary ChatGPT web chats; broader monitoring depends on the plan and supported reporting interfaces.

Needs extra setup: Telemetry & reporting setup
Custom usage monitoring

Standard usage reports are available. For custom monitoring of Claude Code, configure OpenTelemetry centrally and connect a collector, storage and reporting tools. Build the dashboards, alert rules and notification integrations you need. Available events and prompt content depend on export settings and permissions. Claude Code telemetry does not cover every Claude surface; broader monitoring depends on the plan and supported reporting interfaces.

Out of the box: Custom analytics & alerts
Configured in the platform
Platform usage monitoring

Use platform usage records to configure custom reports, monitoring rules and alerts for connected AI tools. Review users, models, token usage, spending and gateway checks, and notify the responsible team when a configured condition matches. Shared platform monitoring reduces the need to assemble a separate telemetry and reporting stack. Coverage follows the tools and traffic connected to the platform; custom reports and notification channels still need configuration.

Live
Grow
Courses & exams
Out of the box: Separate academy
OpenAI Academy

Take courses and complete knowledge assessments through OpenAI Academy. Course learning and credentials are offered through a separate academy, rather than a company learning system built into the Business workspace.

Out of the box: Separate academy
Anthropic Academy

Take courses and complete available assessments and certificates through Anthropic learning resources. These are offered through a separate academy, rather than a company learning system built into the Team workspace.

Out of the box: Inside the platform
Built-in learning

Take courses and exams inside the platform, using a company learning program connected to the tools employees use at work. Track course completion and exam results through the platform.

Live
AI onboarding & coaching
Needs extra setup: Custom setup
Company-specific coaching

Use courses and reusable AI assistants for learning. A company-specific program that combines onboarding agents, assessment of everyday tool use and ongoing progress tracking needs additional configuration and integration. Standard product analytics are not the same as a personalized employee learning program.

Needs extra setup: Custom setup
Company-specific coaching

Use courses and reusable AI assistants for learning. A company-specific program that combines onboarding agents, assessment of everyday tool use and ongoing progress tracking needs additional configuration and integration. Standard product analytics are not the same as a personalized employee learning program.

Out of the box: Built-in coaches &
Progress tracking
AI onboarding and coaching

The platform uses activity from connected tools and the tasks employees complete to assess AI skills and track progress. Built-in onboarding and coaching agents help employees get started, learn tools and build solutions, with suggested next steps for their development. Assessment follows configured criteria and available activity; it measures AI proficiency, not overall employee performance.

Live

Start with one valuable function

See the platform in your own environment.

Choose one function, connect the minimum systems and agree the baseline before expanding.