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

# Resolve MCP Server

Use ResolveAI **as** an MCP server from your coding agent or any MCP client — chat with Resolve, run investigations, and query telemetry as native tool calls.

> Want Resolve to call tools on **another** MCP server instead? See [MCP](/mcp-integrations) under Setup & Integrations.

## Easiest: Install a Plugin

If you use Claude Code, Codex, or Cursor, install the Resolve plugin. It wires up the MCP connection and signs you in with OAuth in one step — nothing to configure by hand. See [AI Assistant Plugins](/ai-assistant-plugins).

## Connect Any MCP Client Directly

For any other MCP client, point it at Resolve's MCP endpoint:

```text theme={null}
https://app0.resolve.ai/mcp/v2
```

Transport: [Streamable HTTP](https://modelcontextprotocol.io/specification/2025-03-26/basic/transports#streamable-http) (stateless — no session management required).

Authenticate one of two ways:

* **OAuth** — clients that support it (like Claude Code and Codex) discover Resolve's OAuth automatically and prompt you to sign in on first connect. Nothing else to set up.
* **API token** — create a personal token at **User menu > Personal API Tokens** (or an organization token at **Admin > API Tokens** for shared/service use), then send it as `Authorization: Bearer <token>`.

Already using the old `/mcp` endpoint? See [Migrating from the MCP Beta](/migrating-from-the-mcp-beta) to switch over.

## Tools

These tools are available to any MCP client connected at `/mcp/v2`. The plugin's skills are built on top of them.

| Tool                  | Description                                                                                                                                           |
| :-------------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------- |
| `ask`                 | Send a question or request to Resolve in a new or existing chat or investigation.                                                                     |
| `start_investigation` | Kick off a deep root-cause investigation from a problem description.                                                                                  |
| `steer_investigation` | Send a finding, hypothesis, or directive into an active investigation.                                                                                |
| `get_investigation`   | Get the full state of an investigation (report, theories, alerts, mitigations, [resolution timing](/resolve-rest-api#resolution-timing-mttrmetrics)). |
| `list_investigations` | List investigations by team, run type, use case, label, and time range.                                                                               |
| `get_chat`            | Get the messages and current state of a chat.                                                                                                         |
| `list_chats`          | List recent chats.                                                                                                                                    |
| `list_alerts`         | List and filter alerts.                                                                                                                               |
| `read_file`           | Read a file path returned by Resolve (cited evidence, report sections).                                                                               |
| `upload_attachment`   | Upload an attachment for use in a chat or investigation.                                                                                              |
| `submit_feedback`     | Submit feedback on an investigation, chat answer, or the product.                                                                                     |

## Live Progress

`ask`, `start_investigation`, and `steer_investigation` may return a `stream_command`: a self-contained `curl` for following live progress. The host agent runs it as a background process and reads stdout until the operation completes. The command carries a short-lived scoped watch token, so no API key is needed.

## Usage and Billing

`ask` handles both new chats and follow-ups. It may start a new billable chat, or continue an existing one, depending on context. Your MCP client decides which tool to call; to be sure you're following up on an existing chat, ask which chat you're in and continue there. Chats carry across platforms (e.g., start one in Slack and continue in Codex).

`start_investigation` kicks off a new billable deep investigation. You can also ask follow-up questions or steer an existing investigation from the ResolveAI plugin or MCP; if it wasn't initiated from the ResolveAI plugin or MCP, join it first.