Packs & MCP

Configure by talking (MCP)

Connect the Turtle AI Coworker MCP connector to Claude, ChatGPT or Cursor and build, install and manage your Turtle AI Coworker workspace by describing what you want in a chat.

The Turtle AI Coworker MCP connector lets an AI assistant you already use, such as Claude, ChatGPT or Cursor, work on your Turtle AI Coworker workspace for you. MCP (Model Context Protocol) is the standard these assistants use to call outside tools. Once connected, you can say "install the Sales Engine pack" or "add a Priority column to my Leads table" and the assistant does it, step by step, in your workspace.

Use it when describing a setup is faster than clicking through it: building from a process document, wiring many pieces at once, auditing what exists, or making the same change across several workers. The dashboard is still home for monitoring: results, run history, costs, approvals and billing live there.

How it works

You askin Claude · ChatGPT · Cursor
ConnectorTurtle AI Coworker MCP server
Your permissionsrole · org · workspace
Workspace changessame as the app
Your assistant turns a request into calls to the Turtle AI Coworker MCP server. Each call runs as you, in the organization and workspace you chose, through the same actions the app uses.

Step by step

  1. Get your connector URL

    The connector URL is the Turtle AI Coworker MCP endpoint with your credentials added as URL parameters:api_key, user_email for your Turtle AI Coworker account, and optionallyorg_id if you belong to more than one organization. The exact format is on theConnect to Claude page.

  2. Add it to your assistant

    In Claude, open Settings, then Connectors, then Add custom connector, paste the URL and save. In ChatGPT, Cursor or another MCP client, add the same URL wherever that client registers a custom MCP server; the menu differs by client. Start a new chat once it is added.

  3. Choose the active organization

    Say "Set up my Turtle AI Coworker workspace" or "List my organizations". The assistant lists your organizations with your role in each and which one is active. If you belong to one organization it is selected for you. If you belong to several, tell it which to use; it sets that organization as active and every later call stays inside it, with no need to edit the URL. An org_id in the URL still takes priority if you set one.

  4. Pick a workspace and see what is there

    Ask it to list your workspaces and pick one, or create a new one. Then ask "What's in this workspace?" for a snapshot of every agent, team, AI employee, table and knowledge base, and "What's left to set up?" for a checklist of what is missing or misconfigured, such as no model connected or a tool not yet authorized. Each item names the next action, and the assistant can work through them with you.

  5. Build or install

    Describe what you need. The assistant can search the template and pack gallery and install a match, then apply your changes to the installed copy. Or it can build from scratch: tables first, then knowledge bases, agents, teams and AI employees, wiring tools, table permissions and triggers as it goes. It confirms its plan with you before creating anything, and asks before deleting.

  6. Connect tools in your browser

    When you ask to connect Gmail, Slack, HubSpot or another app, the assistant gives you a secure link. Open it in your browser and authorize there. The assistant cannot complete the sign-in for you, by design. It then checks the connection and wires the tool to the workers that need it.

Example requests

  • "Set up my Turtle AI Coworker workspace."
  • "I need something for recruiting outreach. What do you have?"
  • "Install the Sales Engine solution pack and rename the SDR to Maya."
  • "Connect my Gmail."
  • "What's left to set up?"
  • "Create a table called Leads with Name, Company, Email and a Status select of New, Contacted and Qualified."
  • "Give the Lead Scorer agent read and update access to the Leads table, but not delete."
  • "Every Monday at 9am London time, have my AI employee send me a pipeline summary."
  • "List every agent in this workspace that has no tools attached."
  • "Add a policy that requires approval before any email is sent."

What the connector can do

The connector groups its actions by area. You do not need these names; they help you know what to ask for.

AreaTypeWhat it does
organizationsscopeList your organizations with your role, and set the active one.
authscopeWho you are, and list or create workspaces.
workspaceoverviewOne snapshot of every agent, team, AI employee, table and knowledge base in a workspace.
next_stepschecklistWhat is missing or misconfigured, with the next action for each item.
templatesgallerySearch the gallery, install a template, or save a live agent, team, AI employee or table as a template.
solution_packspacksPreview, check and install packs, or bundle resources you built into a pack of your own.
agents, agent_config, agent_runsworkersCreate and edit sequential agents, attach tools and knowledge bases, run them and read run results.
agent_tool_triggersautomationRun an agent when a connected tool sends an event, such as a new email.
teams, team_config, team_chat, team_sessionsworkersCreate agentic teams, add members, attach tools, knowledge bases and tables, chat with a team and review sessions.
ai_employees and relatedworkersCreate AI employees, grant access to agents, tables, knowledge bases and tools, chat, manage tasks, schedules, event triggers and memory.
tables, table_triggers, table_agent_accessdataTables, columns and rows, row triggers and schedules, and per-agent Read, Create, Update and Delete access.
knowledge_basesdataCreate knowledge bases and add text, web pages or files.
integrations, tools_catalogconnectionsBrowse the tool catalog, get a sign-in link, and check a connection.
llm_providersmodelsSee which model providers and models your organization can use.
policies, budgets, circuit_breakersgovernancePolicy rules, spend caps and the thresholds at which runs are throttled or stopped.
get_instructionsguidesBuilt-in guides the assistant reads before it builds: setup, managing, packs and each worker type.

Related