MODEL CONTEXT PROTOCOL

Connect your AI client to Bizcraw.

Use the Bizcraw MCP server to let a compatible LLM read approved workspace context and call supported sales tools within your permissions.

Production MCP endpointhttps://bizcraw.com/api/mcp

Use OAuth when the client supports it. Otherwise, create a scoped access token in your Bizcraw workspace.

Standard HTTP transport

Connect through the production HTTPS endpoint from clients that support remote MCP servers.

Scoped credentials

Choose only the read or write permissions the client needs and rotate or revoke a connection at any time.

Workspace boundaries

Requests remain tied to the selected workspace, current plan access and server-side authorization checks.

01

Before connecting

Create or join a Bizcraw workspace whose plan includes MCP. A workspace owner or admin opens Workspace → MCP, checks that the service is enabled and creates the connection.

  • Use a descriptive connection name for each client.
  • Grant the minimum scopes required for the intended workflow.
  • Choose an expiry date and rotate credentials periodically.
02

Connect with OAuth

For ChatGPT and any client that supports remote MCP OAuth, add a custom MCP app and enter the production server URL. The client redirects to Bizcraw so the user can sign in, review requested scopes and authorize the workspace.

SERVER URL
https://bizcraw.com/api/mcp

OAuth is the preferred path when supported because the user reviews access in Bizcraw and can revoke it later without editing local configuration.

03

Connect with an access token

In Bizcraw, create a connection and copy the token when it is shown. The token is the value that replaces <YOUR_TOKEN>; it is not your password, workspace ID or an API key from another provider.

Claude, Cursor and compatible JSON clients

{
  "mcpServers": {
    "bizcraw": {
      "type": "http",
      "url": "https://bizcraw.com/api/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_TOKEN>"
      }
    }
  }
}

Codex

[mcp_servers.bizcraw]
url = "https://bizcraw.com/api/mcp"
bearer_token_env_var = "BIZCRAW_MCP_TOKEN"

Set BIZCRAW_MCP_TOKEN in the local environment used to start Codex. Do not commit the token to a repository or paste it into ordinary prompts.

04

Verify the connection

Restart or refresh the client, then ask: What Bizcraw workspace am I connected to? A working connection should call get_workspace and the Bizcraw MCP page should show a recent Last used time.

  • A 401 response means the token is missing, expired or revoked.
  • A 403 response can mean the plan, workspace state or requested scope does not permit the operation.
  • If tools are absent, confirm the platform rollout and read/write controls are enabled.
05

Permissions and safety

Bizcraw exposes workspace, CRM, data, knowledge, quote and scraping tools according to the scopes granted to the connection. Read tools retrieve only workspace-scoped data. Supported write tools use server-side validation, audit records and idempotency controls.

Approval remains inside Bizcraw

Actions that can incur provider usage or create an external effect may wait for owner or admin approval. Review pending requests from Workspace → MCP.

QUESTIONS

MCP connection answers.

What is the Bizcraw MCP server?+

It is a secure interface that lets compatible AI clients use approved Bizcraw tools against the workspace connected by the user.

What replaces YOUR_TOKEN?+

Create an MCP connection inside Bizcraw and copy the access token shown once. Paste that token in the client configuration or store it in the named environment variable.

Can an AI client access every workspace?+

No. Each connection belongs to one workspace, carries explicit scopes and is subject to the user's role, plan access and server controls.

Can the client make changes automatically?+

Only granted write tools are available. Paid or externally meaningful operations can create an approval request that a workspace owner or admin must review in Bizcraw.

CONNECT BIZCRAW

Give your AI client controlled access to the workspace where sales work happens.

Start with read scopes, verify the connection, then add write permissions only when the workflow requires them.

Create workspace