OpenClaw and Other AI Agents

OpenClaw on the Pro plan

On the Pro plan, Customermates does not expect every team to wire an AI agent setup from scratch. Instead, the product provisions a preconfigured, sandboxed OpenClaw container for each user, so the agent runtime is already prepared for CRM-oriented usage.

That gives you an agent environment that is ready to be connected to your own provider keys and CRM workflows, while still keeping execution isolated per user instead of running a shared agent process across the workspace.

How the setup works

The OpenClaw container is prepared by Customermates as a managed runtime. Access to CRM data remains user-scoped, which means the agent works with the permissions of the user it belongs to instead of receiving broad workspace-wide access by default.

In practice, the setup looks like this:

1. The user is on the Pro plan.
2. Customermates provisions a sandboxed OpenClaw container for that user.
3. The user adds their own provider credentials such as OpenAI or Anthropic keys.
4. The agent works against Customermates with user-scoped permissions.

Why MCP still matters

Even though the OpenClaw container is preconfigured by Customermates, the integration approach still follows the Model Context Protocol (MCP). That gives the agent a tool-native way to discover CRM capabilities, execute actions, and recover from validation issues without forcing the runtime to operate against the full raw API surface.

Use OpenAPI 3.1.0 alongside it only when you need exact endpoint payloads, schema details, or lower-level transport validation.

Available container actions

Once the container exists, Customermates exposes three practical actions for managing it:

• Open Control UI to open the agent control interface in a separate window
• Add Environment Variable to add or update runtime configuration for the container
• Reset Agent to fully reset the agent state when you want a clean start

What users typically configure

After provisioning, the most common next step is to add provider credentials such as OpenAI or Anthropic keys. From there, teams usually add environment variables only when they need additional runtime configuration for prompts, tools, or external services.

That means the user does not start by assembling the full container architecture themselves. The core runtime is already there, and they mainly supply credentials and environment-specific configuration.

Skills and Orthogonal

Once the container is provisioned, the next practical step is usually the Skills page. That is where users can browse reusable capabilities that can be added to the agent for more specific workflows.

Some of those skills come from Orthogonal. In this context, Orthogonal is an external skill source that provides reusable agent capabilities which can be installed into the OpenClaw runtime and then combined with Customermates workflows.

Security and runtime boundaries

The key design principle stays the same throughout the setup: CRM access remains user-scoped and the OpenClaw runtime stays sandboxed. That separation matters because it keeps agent execution isolated per user while still allowing the agent to work with live CRM context.

For the broader runtime model behind that setup, continue with Architecture & Security. For event-driven follow-up flows after agent actions, pair this with Webhooks & Events.

OpenClaw setup in the product