> ## Documentation Index
> Fetch the complete documentation index at: https://help.bodharouteplanner.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP FAQ

> Frequently asked questions about the Bodha Route Planner MCP server.

# Frequently Asked Questions

> **What is MCP?**
>
> MCP stands for **Model Context Protocol** — an open standard that lets AI assistants connect to external tools and data sources. It's supported by Claude Desktop, Cursor, Windsurf, VS Code Copilot, and a growing number of AI clients. Think of it as giving your AI assistant the ability to look things up in your business systems while it's talking to you.

> **Which AI assistants work with this?**
>
> Any AI assistant that supports **MCP with Streamable HTTP transport**. Currently confirmed:
>
> * **Claude Desktop** (by Anthropic)
> * **Cursor** (AI code editor)
> * **Windsurf** (by Codeium)
> * **VS Code** with Copilot MCP extensions
> * Any other MCP-compatible client that supports Streamable HTTP
>
> The protocol is open and growing. New AI assistants add MCP support regularly.

> **Can the AI modify my routes or data?**
>
> **No.** The MCP server is strictly **read-only**. AI assistants can query your routes, stops, drivers, orders, analytics, and depots — but they cannot create, edit, or delete anything. Write capabilities are planned for a future release and will require explicit write scopes.

> **Does this cost extra?**
>
> No per-call charges. MCP access is **included** in your Pro or Enterprise plan at no additional cost. Use it as much as you want — we don't meter individual tool calls.
>
> | Plan | MCP access |
> | - | - |
> | **Starter** | Not available |
> | **Pro** | 1 API key, all read scopes |
> | **Enterprise** | Unlimited keys, all scopes, extended audit log |

> **Does the AI assistant store my data?**
>
> Bodha Route Planner sends data to the AI assistant in response to tool calls during your conversation. Each AI provider has their own data handling and retention policies. Bodha Route Planner does not control how the AI assistant processes data after it's received. Consult your AI provider's privacy policy for details:
>
> * [Anthropic (Claude)](https://www.anthropic.com/privacy)
> * [Cursor](https://cursor.com/privacy)
> * [Codeium (Windsurf)](https://codeium.com/privacy-policy)

> **Can other organizations see my data?**
>
> **Absolutely not.** Every query is scoped to your organization (tenant) at the database level. Your API key is tied to your organization, and every data request includes your tenant ID as a mandatory filter. There is no API endpoint or tool call that can return another organization's data.

> **What happens if my API key is leaked?**
>
> 1. **Revoke it immediately** — Go to **Settings → API Keys** and click **Revoke** next to the compromised key. It stops working on the very next request.
> 2. **Check the audit log** — Review **Settings → API Keys → Activity** to see if the key was used by anyone unexpected.
> 3. **Create a new key** — Generate a replacement and update your AI assistant's config.
>
> Because the MCP server is read-only, a leaked key cannot modify your data — but it could expose route and stop details to an unauthorized reader. Revoke immediately.

> **Why is the first request slow?**
>
> The MCP server runs on serverless infrastructure (AWS Lambda). The first request after a period of inactivity triggers a "cold start" — the server needs 2–5 seconds to initialize. Subsequent requests in the same session are much faster (typically under 200ms). This is normal and expected.

> **Is there a rate limit?**
>
> Yes — **60 read requests per minute per key**, with a hard ceiling of **10 requests per second**. Normal conversational use (asking questions one at a time in an AI chat) will never hit these limits. They exist to prevent misconfigured scripts or runaway automations.

> **Can I use this for automated reporting?**
>
> The MCP server is designed for conversational AI interactions, not batch automation. For automated reporting, scheduled data pulls, or integration pipelines, use the [REST API](https://help.bodharouteplanner.com/api-reference/introduction.md) instead — it's built for programmatic access with higher rate limits and webhook support.
>
> That said, many teams use the MCP server for **ad-hoc reporting** through their AI assistant — asking for summaries, comparisons, and insights on demand. This works great and doesn't hit rate limits.

> **Can multiple people in my organization use it?**
>
> On the **Pro** plan, you get 1 API key — one person can use it at a time (or share the key, but that makes audit logging less useful). On **Enterprise**, you get unlimited keys, so each team member can have their own key with their own scopes and audit trail.

> **Will write capabilities be added?**
>
> Yes — a future release will add write tools (create orders, update stop status, dispatch routes) behind explicit write scopes. Write scopes will be opt-in and separate from read scopes, so existing read-only keys won't gain write access automatically.

> **Do I need to install anything locally?**
>
> **No.** The MCP server is fully hosted on Bodha Route Planner's infrastructure. You don't need to install Node.js, Docker, npm packages, or run any local server. Just paste a JSON config into your AI assistant and you're connected.

> **Can I use this with Claude Code (CLI)?**
>
> Yes. Claude Code supports MCP servers. Add the Bodha Route Planner config to your Claude Code MCP settings and you can query fleet data from the terminal while coding.

***

## Still have questions?

Contact [support@bodharouteplanner.com](mailto:support@bodharouteplanner.com) — we're happy to help with setup, scoping, or any technical questions about the MCP server.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.