> For the complete documentation index, see [llms.txt](https://integrations.impact.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://integrations.impact.com/ai-solutions/mcp-quick-start/mcp.impact.com-faq.md).
# mcp.impact.com FAQ
mcp.impact.com/mcp is impact.com's public Model Context Protocol (MCP) endpoint, now available in Open Beta. Connect a compatible AI client to your impact.com account to use a set of supported tools through conversational workflows. This FAQ answers common questions about what the beta includes, how setup and access work, and how security is handled.
## General
What is mcp.impact.com?
`mcp.impact.com` is impact.com's public MCP endpoint. It allows an MCP-compatible AI client to connect to an impact.com account so supported impact.com tools and data can be used through conversational workflows.
In practical terms, this helps customers work with selected impact.com capabilities through compatible AI tools without needing to build a custom integration from scratch.
Where can I get started?
The best starting point is [the MCP quick start guide](https://integrations.impact.com/ai-solutions/mcp-quick-start). That documentation explains how to enable MCP access, which clients are supported, which tools are available, and how to complete setup.
Does this replace the impact.com platform or user interface?
No. The MCP service extends access to selected impact.com capabilities through compatible AI tools. It does not replace the core impact.com platform or its standard workflows.
Which AI clients are supported?
The service is designed for MCP-compatible clients. Claude Code, Cursor, and VS Code are currently supported.
For setup guidance, see the [quick start documentation](https://integrations.impact.com/ai-solutions/mcp-quick-start).
Which tools are supported?
For the latest supported tool list, setup, and support information, visit [integrations.impact.com](https://integrations.impact.com).
What should I do if setup fails?
First, confirm that MCP access is enabled for your account and that you are following the documented setup flow for your chosen client.
If the issue continues, contact your normal impact.com support channel or customer contact for assistance.
Where is the MCP hosted?
The MCP endpoint is `https://mcp.impact.com/mcp`. Authentication via OAuth flows through `app.impact.com` issuing a scoped JWT to the AI client. Access is restricted to the MCP endpoint only.
## Beta
What does "Beta" mean?
Beta means the service is available for customer use, with an intentionally focused launch scope. The beta includes a curated set of supported tools, setup documentation, and an active feedback loop as the experience continues to improve.
The beta can be used for supported scenarios today, while some capabilities, documentation, and operational maturity continue to evolve over time.
Who is this beta for?
This beta is designed for customers who want to connect their own MCP-compatible client to impact.com and use supported tools in an AI-driven workflow.
It is especially useful for technical users, developers, solutions teams, and operational users who are comfortable following a guided setup process.
What can I do with the beta?
You can use a curated set of supported, customer-facing MCP tools that are available as part of the beta. These tools are intended to support practical workflows through an MCP-compatible client.
Will all MCP tools be available at launch?
No. The beta includes a curated set of supported tools rather than the full possible MCP tool set.
This focused approach helps ensure a better experience for the tools included in the initial release.
What are the main limitations of the beta?
The main limitations are a curated tool set, evolving documentation, and the normal changes that come with a beta-stage release. Some capabilities that may be valuable in the future are not included in the initial launch scope.
## MCP access
How do I get access?
Access is enabled through an account-level MCP setting. Once MCP access is turned on for the account, the setup instructions at [integrations.impact.com](https://integrations.impact.com) can be followed to connect a compatible client.
Are API keys required to use MCP?
No. The primary experience uses a one-time OAuth authorization flow. Users sign in with their existing impact.com credentials and grant consent to the supported AI client.
This makes setup simpler and removes the need to create and manage separate API keys.
Can the MCP access other impact.com endpoints?
The token issued during the OAuth authorization process is restricted to the MCP endpoint only.
How can I switch to a different account if I have multiple accounts?
For the beta, switching between accounts requires a manual workaround.
Steps:
1. Disconnect the impact.com MCP connection within the AI client.
2. Log into [app.impact.com](https://app.impact.com/).
3. Switch the account in [app.impact.com](https://app.impact.com) to the account to be connected in the MCP.
4. Reconnect the impact.com MCP connection.
1. The user will be guided through an OAuth login flow.
2. Once successfully logged in, the available MCP tools will be listed in the AI client.
In the future, an MCP tool will be created to help facilitate the account switch. The OAuth login flow will also be enhanced to enable switching to a chosen account at the time of connecting the MCP.
Does enabling MCP give every user the same level of access?
No. MCP access is enabled at the account level, but each user can only access what their existing [impact.com](http://www.impact.com/) permissions allow.
Does disabling MCP affect the whole account?
Yes. This setting applies to the entire account, not to individual users.
What I do if users still cannot connect after MCP access is enabled?
Confirm the setting was saved successfully, then ask the user to try the connection again using the documented setup flow for the supported AI client. If the issue continues, follow the normal [impact.com](http://impact.com) support process.
## Manage tool permissions
How do I manage tool permissions in Cursor?
Connecting mcp.impact.com authorizes Cursor to call supported tools within your existing impact.com permissions. Separately, Cursor may ask you to approve each tool call, or remember that approval. Those prompts are controlled by Cursor, not by impact.com.
#### **Steps**
1. When the agent wants to use an impact.com MCP tool, review the prompt and approve or deny.
2. To change how often you are asked: open **Cursor Settings → Agents → Approvals & Execution** and choose a run mode (for example, **Auto-review** or **Allowlist**).
3. To let trusted MCP tools run without prompting, add them to the MCP allowlist from that settings area, or choose **Add to allowlist** when prompted.
4. Optionally, enable or disable an MCP server from the tools list in chat.
1. Select **Customize**, then the tools list at the top of the chat panel.
Denying a prompt blocks that call in Cursor. It does not change your impact.com account permissions or the account-level MCP enablement setting.
How do I manage tool permissions in VS Code?
Connecting mcp.impact.com authorizes VS Code to call supported tools within your existing impact.com permissions. Separately, VS Code may ask you to approve each tool call, or remember that approval. Those prompts are controlled by VS Code, not by impact.com.
#### Steps
1. When Copilot Chat or the agent needs an MCP tool, review the confirmation dialog (tool name and parameters) and approve or deny.
2. To manage saved approvals: open the Command Palette and run **Chat: Manage Tool Approval**.
1. Tools are grouped by MCP server. Adjust per tool, or trust all tools from a specific server.
3. To change session autonomy: use the permissions dropdown in the chat input (**Default Approvals**, **Bypass Approvals**, or **Autopilot**).
1. Prefer **Default Approvals** for day-to-day use.
4. To clear saved approvals: open the Command Palette and run **Chat: Reset Tool Confirmations**.
Denying a prompt blocks that call in VS Code. It doesn't change your impact.com account permissions or the account-level MCP enablement setting.
How do I manage tool permissions in Claude Code?
Connecting mcp.impact.com authorizes Claude Code to call supported tools within your existing impact.com permissions. Separately, Claude Code may ask you to approve each tool call, or remember that approval. Those prompts are controlled by Claude Code, not by impact.com.
#### **Steps**
1. When Claude Code prompts to use an impact.com MCP tool, approve once, or choose the option to remember the choice when offered.
2. To review or change rules: run `/permissions` and set **Allow**, **Ask**, or **Deny** for the relevant MCP tools.
3. To see connected MCP servers and tools: run `/mcp`.
4. Permission modes (for example, Manual, Auto, or Bypass) change how often you are prompted. Use bypass modes only in environments you trust.
Denying a prompt blocks that call in Claude Code. It does not change your impact.com account permissions or the account-level MCP enablement setting.
Why does my AI client keep asking me to approve impact.com tools?
Your AI client controls whether tool calls require approval. If you are prompted every time, check the client’s permission or run mode settings and, where available, add trusted impact.com MCP tools to an allowlist or choose “always allow” / “don’t ask again” when offered.
Use the client-specific steps above for Cursor, VS Code, or Claude Code.
These prompts are separate from your impact.com permissions. Approving a tool in your AI client does not grant access beyond what your impact.com role already allows.
## Security
How does authentication work?
The beta uses OAuth with a standard authorization flow for compatible MCP clients. When a client connects, the user is redirected to sign in with their existing impact.com account and approve access.
Once that step is complete, the client can make authenticated requests on the user's behalf within the supported scope.
Is my data secure?
Yes. The beta uses OAuth-based authentication and explicit user consent. The MCP service is designed as a controlled layer in front of existing impact.com capabilities rather than providing unrestricted access.
As with any beta, we recommend using the service within the documented scope and supported setup paths.
## Troubleshooting
Why is my Cursor tools list out of date after I reconnect?
Cursor can cache the MCP tools list, so it may not show the latest tools after you reconnect.
**Fix:** Quit Cursor completely and reopen it. When you view the tools in the MCP list, the list should refresh automatically.
This is a known Cursor client behavior, not an impact.com MCP issue.
Why does my Claude MCP connection drop after a few minutes?
impact.com MCP access tokens are short-lived (about 5 minutes). Claude normally refreshes them silently using a longer-lived refresh token.
If a refresh fails, for example because the refresh token was revoked, or your network blocks `app.impact.com/oauth2/token` , the connection can drop mid-session.
**Fix:** Restart the Claude MCP connection (or **Disconnect** → **Connect** for the Claude Desktop connector) and complete the login again if prompted. If drops keep happening, check that your network can reach `https://app.impact.com` , including the token endpoint.
How do I re-authenticate Claude with the impact.com MCP?
Use the path that matches how you connected:
* **Claude Desktop connector UI:**
* Open **Customize → Connectors**, select the impact connector, then **Disconnect** and **Connect** again.
* **Claude Code or Claude Desktop JSON /** `mcp-remote` **config:**
* Delete the cached tokens with `rm -rf ~/.mcp-auth/`, then restart Claude and run the OAuth flow again (for Claude Code, use `/mcp` and select `impact`).
After config changes in the JSON path, quit Claude completely. Fully exit the app, don't just close the window, then reopen it.
If the server does not appear when using the JSON / `mcp-remote` path, confirm Node.js 18+ is installed and `npx` is on your `PATH` . Claude uses `npx` to start `mcp-remote`.
Why does Claude say “Incompatible auth server: does not support dynamic client registration”?
impact.com MCP doesn't support Dynamic Client Registration (DCR) as part of the open beta. Claude must use a pre-registered client configuration rather than registering a client on the fly.
If you see this error, follow the Claude setup steps in [MCP for Claude Desktop & Claude Code](https://integrations.impact.com/ai-solutions/mcp-quick-start/mcp-for-claude-desktop-and-claude-code). Use the path that provides the published impact.com OAuth Client ID (Claude Desktop connector UI, or the JSON / static-client configuration).
For Claude Desktop, prefer the built-in connector UI when available. This is the simplest supported path during the open beta.
I got an authorization error when connecting, but no sign-in window appeared. What should I do?
When you select **Connect**, your AI client opens a browser window or tab so you can sign in to [impact.com](http://impact.com/) and approve access.
* If you have many tabs open, or the browser blocks pop-ups, that window can sit behind other tabs.
* Cursor or Claude may then show an error such as *Authorization with* [impact.com](http://impact.com/) *failed* even though your account is eligible and MCP Access is on.
* This is not usually a permissions problem on your [impact.com](http://impact.com/) account.
**What to try**
* Look for a new browser tab or pop-up for [impact.com](http://impact.com/) sign-in, and complete it.
* Close extra tabs, or bring your browser to the front, then select **Connect** again.
* If pop-ups are blocked, allow them for your AI client and retry.
* Completing **Connect** in a less crowded browser session (or Claude Desktop) can make the sign-in window easier to see.
If you still cannot complete login after a visible [impact.com](http://impact.com/) window appears, contact your [impact.com](http://impact.com/) support channel and include the exact error text.
## Glossary of terms
| Term | Definition |
| --------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| MCP | Abbreviation of Model Context Protocol. MCP is a standard way for AI clients to connect to external services, tools, and data sources. For mcp.impact.com, it enables compatible AI clients to interact with selected impact.com capabilities. |
| Supported AI client / supported AI tool / MCP-compatible client | An AI application or development tool that can connect to an MCP server and is supported for use with mcp.impact.com.
Examples currently include Claude Code, Cursor, and VS Code. Use the documented setup instructions for supported clients.
|
| MCP tool | A specific capability exposed through the MCP service that an AI client can call on behalf of an authenticated user.
MCP tools are limited to the curated, customer-facing set made available for the beta.
|
| Authentication | The process of confirming who the user is. In this beta, users authenticate with their existing impact.com credentials during the OAuth flow. |
| Authorization | The process of determining what an authenticated user is allowed to access or do. A user's impact.com permissions determine whether they can use a specific MCP tool or capability. |
| OAuth | An authorization framework that lets a user grant an AI client permission to access supported impact.com capabilities without sharing separate API keys. The beta uses a one-time OAuth authorization flow with explicit user consent. |
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.
Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:
```
GET https://integrations.impact.com/ai-solutions/mcp-quick-start/mcp.impact.com-faq.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.