MCP tools for agents - Mapping API Docs

The mapping.travel MCP server exposes a read + chat subset of the agents surface so you can manage agents from your favourite LLM client. All new agent tools require a Pro Max subscription — the server returns a typed 403 on non-Pro-Max workspaces.

New tools

Tool	Mutating?	Description
`list_agents`	No	All agents in the workspace.
`get_agent`	No	Single agent with current status, scope, rule-set summary.
`list_agent_runs`	No	Recent runs for an agent (filter by status, since).
`get_agent_run`	No	One run with optional step inclusion.
`list_agent_pending_items`	No	Open pending items across one agent or all.
`list_agent_threads`	No	Conversation threads owned by the calling user.
`chat_with_agent`	Yes (creates message + cycle)	Send a message, get the reply. Non-streaming over MCP.
`pause_agent`	Yes	Idempotent.
`resume_agent`	Yes	Idempotent.
`get_agent_costs`	No	MTD vs cap, per-day breakdown.
`list_agent_data_sources`	No	All configured sources + sinks.

chat_with_agent is not idempotent — each call creates a thread message and (potentially) a cycle.

Per-client setup

You need an API key from Developers → MCP in the app. The setup is identical to the existing MCP setup; the new agent tools appear automatically once your workspace is on Pro Max.

Claude Desktop
Cursor
Continue
MCP Inspector

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "mapping-travel": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote@0.1.37",
        "https://mcp.mapping.travel/mcp",
        "--header",
        "X-API-Key: mt_mcp_live_..."
      ]
    }
  }
}

Restart Claude Desktop. You should see ~29 tools under the wrench icon (18 existing + 11 new agent tools).Try: “List my agents and show me the open pending items for the first one.”

Edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "mapping-travel": {
      "url": "https://mcp.mapping.travel/mcp",
      "headers": { "X-API-Key": "mt_mcp_live_..." }
    }
  }
}

Settings → MCP should show 29 tools.

Edit ~/.continue/config.json:

{
  "experimental": {
    "modelContextProtocolServers": [
      {
        "transport": {
          "type": "streamable-http",
          "url": "https://mcp.mapping.travel/mcp",
          "headers": { "X-API-Key": "mt_mcp_live_..." }
        }
      }
    ]
  }
}

For a quick visual check without configuring a long-running client:

npx @modelcontextprotocol/inspector

Pick Streamable HTTP, URL https://mcp.mapping.travel/mcp, header X-API-Key: mt_mcp_live_..., click Connect. You can invoke any of the 29 tools from the form.

Tier gating

If your workspace is not on Pro Max, calls to any list_agent_*, chat_with_agent, pause_agent, resume_agent, or get_agent_costs return:

{
  "type": "https://mapping.travel/errors/plan-required",
  "title": "Pro Max plan required",
  "status": 403,
  "requiredPlan": "PROFESSIONAL_MAX",
  "currentPlan": "PROFESSIONAL"
}

The non-agent tools (mapping, search, inventory) continue to work normally.

Streaming

chat_with_agent is non-streaming over MCP — transports behave inconsistently across clients, and most TUI usage prefers complete replies with a summary of the tool calls the agent made. Streaming chat stays UI-only (the app uses SSE).

Authorization

Role	Can do
`OWNER`, `ADMIN`	Everything above plus pause/resume.
`MEMBER`	Read + `chat_with_agent` (rate-capped).

Troubleshooting

See MCP troubleshooting — the agent tools share the same transport, so any connectivity issue affects all 29 tools uniformly.

Agents overview — why you’d want to chat with an agent.
MCP connect — the original per-client connection guide.

​New tools

​Per-client setup

​Tier gating

​Streaming

​Authorization

​Troubleshooting

​Related