# Connecting MCP Clients to PipesHub MCP Server
This guide covers how to connect PipesHub's remote MCP server to **Cursor**, **Claude Code**, **Gemini CLI**, **Codex CLI**, **Claude.ai (Web)**, and **LibreChat** using static OAuth credentials or bearer tokens.
PipesHub exposes a remote MCP endpoint over **Streamable HTTP** at `/mcp`. MCP Clients connect to this endpoint directly -- no local npm packages or stdio processes needed.
[](https://youtu.be/VgyND38QZ7w)
**[▶ Watch the 2½-minute demo](https://youtu.be/VgyND38QZ7w)**: Claude Code using this MCP server to answer from Jira, Slack, GitHub and Drive, with each person's own permissions.
> **Coding agent?** Start at [For coding agents](https://docs.pipeshub.com/for-agents.md). Install the skill into the *user's* repo with `npx skills add pipeshub-ai/mcp-server` (see [`skills/pipeshub`](./skills/pipeshub/SKILL.md)) and append the `AGENTS.md` snippet on that page. Listed on the [official MCP registry](https://registry.modelcontextprotocol.io) as `io.github.pipeshub-ai/mcp` and on [Cursor Directory](https://cursor.directory/plugins/mcp-server-3) as PipesHub. The listing files (`plugin.json`, `mcp.json`) default MCP to `http://localhost:3000/mcp` (Docker). Change the URL for a company instance; they contain no secrets. Contributors working in this repository: read [AGENTS.md](./AGENTS.md).
>
> **Claude Code plugin:** `/plugin marketplace add pipeshub-ai/mcp-server`, then `/plugin install pipeshub@pipeshub`, adds the MCP server, the skill, and the `/pipeshub:ask` and `/pipeshub:setup` commands. It connects to `${PIPESHUB_MCP_URL:-https://app.pipeshub.com/mcp}` with `Authorization: Bearer ${PIPESHUB_MCP_TOKEN}`; export both in your shell (self-hosted: set the URL to `https://<your-instance>/mcp`), then restart Claude Code.
>
> **Looking for the tool reference?** See [TOOLS.md](./TOOLS.md) for descriptions, arguments, and a decision guide for each tool the MCP server exposes (`pipeshub_chat`, `pipeshub_search`, `pipeshub_get_record_content`, `pipeshub_download_record`, `pipeshub_directory`, `pipeshub_sources`, `pipeshub_agents`). Searching and reading records is the recommended default; `pipeshub_chat` is optional (see [Leaving `pipeshub_chat` out](#leaving-pipeshub_chat-out)).
>
> **Using QM?** QM cannot attach a third-party MCP endpoint — it is an MCP *server* to its own harness, not a client. Follow [Use PipesHub with QM](./qm/docs/use-with-qm.md). The deployment-layer bundle in [`qm/`](./qm/) gives agents a `pipeshub` command inside their sandbox; this package ships that command as a second bin.
## Prerequisites
- A running PipesHub instance: [PipesHub Cloud](https://app.pipeshub.com) or self-hosted. To self-host, the [local Docker demo](https://docs.pipeshub.com/for-agents-local-demo.md) playbook covers install and first-run — do not scaffold LangChain. First-run (account + LLM) is still in the browser; search 500s until an LLM is configured.
- An OAuth app created in PipesHub (see [Step 1](#step-1-create-an-oauth-app-in-pipeshub))
## Step 1: Create an OAuth App in PipesHub
1. Log in to your PipesHub instance as an admin
2. Navigate to **Settings > Developer Settings > OAuth Apps**
3. Click **Create OAuth App**
4. Fill in the app details:
- **Name**: e.g., `MCP Integration`
- **Redirect URIs**: Add all the redirect URIs for the clients you plan to use:
| Client | Redirect URI |
|---|---|
| **Cursor** | `cursor://anysphere.cursor-mcp/oauth/callback` |
| **Claude Code** | `http://localhost:<PORT>/callback` (e.g., `http://localhost:8080/callback`) |
| **Claude.ai (Web)** | `https://claude.ai/api/mcp/auth_callback` |
| **Gemini CLI** | `http://localhost:7777/oauth/callback` |
| **LibreChat** | `http://localhost:3080/api/mcp/<server-identifier>/oauth/callback` |
> **Important:** The scopes in [`MCP_SCOPES`](https://github.com/pipeshub-ai/pipeshub-ai/blob/main/backend/env.template#L57) must match the scopes granted to your OAuth app — a mismatch will result in an authorization error.
5. Save the app and copy the **Client ID** and **Client Secret**
### Customizing Default Scopes
By default, PipesHub exposes some default scopes in its `/.well-known/oauth-protected-resource/mcp` discovery endpoint. You can customize which scopes are exposed by setting the [`MCP_SCOPES`](https://github.com/pipeshub-ai/pipeshub-ai/blob/main/backend/env.template#L57) environment variable on your PipesHub instance. This is useful for clients like Claude Code that automatically request all exposed scopes.
## Placeholders
Replace these in all configurations below:
| Placeholder | Description | Example |
|---|---|---|
| `PIPESHUB_INSTANCE_URL` | Your PipesHub instance URL | `https://app.pipeshub.com` |
| `YOUR_CLIENT_ID` | OAuth app client ID | `clid_abc123...` |
| `YOUR_CLIENT_SECRET` | OAuth app client secret | `clsec_xyz789...` |
The remote MCP endpoint URL is: `PIPESHUB_INSTANCE_URL/mcp`
---
## Remote MCP Setup
<details>
<summary><strong>Cursor</strong></summary>
Cursor supports static OAuth for remote MCP servers via the `auth` object in `mcp.json`.
### Configuration
Open Cursor Settings > Tools and Integrations > New MCP Server, or edit your project's `.cursor/mcp.json`:
```json
{
"mcpServers": {
"pipeshub": {
"url": "PIPESHUB_INSTANCE_URL/mcp",
"auth": {
"CLIENT_ID": "YOUR_CLIENT_ID",
"CLIENT_SECRET": "YOUR_CLIENT_SECRET",
"scopes": [
"org:read", "org:write", "org:admin",
"user:read", "user:write", "user:invite", "user:delete",
"usergroup:read", "usergroup:write",
"team:read", "team:write",
"kb:read", "kb:write", "kb:delete", "kb:upload",
"semantic:read", "semantic:write", "semantic:delete",
"conversation:read", "conversation:write", "conversation:chat",
"agent:read", "agent:write", "agent:execute",
"connector:read", "connector:write", "connector:sync", "connector:delete",
"config:read", "config:write",
"document:read", "document:write", "document:delete",
"crawl:read", "crawl:write", "crawl:delete"
]
}
}
}
}
```
Cursor will auto-discover the authorization and token endpoints via PipesHub's `/.well-known/oauth-protected-resource/mcp` metadata.
> **Note:** If the `scopes` field is omitted, Cursor fetches `/.well-known/oauth-protected-resource/mcp` and requests **all** `scopes_supported` listed there. To limit access, explicitly list only the scopes you need. You can also control which scopes are exposed server-side — see [Customizing Default Scopes](#customizing-default-scopes).
### Using Environment Variables
Use Cursor's `${env:VAR}` interpolation to keep secrets out of config files:
```json
{
"mcpServers": {
"pipeshub": {
"url": "${env:PIPESHUB_INSTANCE_URL}/mcp",
"auth": {
"CLIENT_ID": "${env:PIPESHUB_CLIENT_ID}",
"CLIENT_SECRET": "${env:PIPESHUB_CLIENT_SECRET}",
"scopes": [
"kb:read", "kb:write",
"semantic:read", "semantic:write",
"conversation:read", "conversation:write", "conversation:chat",
"agent:read", "agent:write", "agent:execute",
"connector:read", "connector:write",
"config:read", "user:read"
]
}
}
}
}
```
### Redirect URI
Cursor uses a fixed redirect URI for all MCP servers:
```
cursor://anysphere.cursor-mcp/oauth/callback
```
Register this as the allowed redirect URI when creating the OAuth app in PipesHub.
### OAuth Login Troubleshooting
If Cursor's internal browser fails to load the OAuth login page, copy the authorization URL from the internal browser and paste it into your normal browser to complete the login flow.
</details>
<details>
<summary><strong>Claude Code</strong></summary>
Claude Code supports remote HTTP MCP servers with static OAuth credentials via `--client-id`, `--client-secret`, and `--callback-port`.
PipesHub exposes discovery at `/.well-known/oauth-protected-resource/mcp`, so Claude Code auto-discovers the authorization and token endpoints.
> **Important:** Claude Code does **not** support configuring specific scopes. It fetches `/.well-known/oauth-protected-resource/mcp`, reads the `scopes_supported` list, and requests **all** of them. Your OAuth app in PipesHub **must have access to all scopes** listed in the discovery endpoint, otherwise the authorization request will fail. To limit the exposed scopes, see [Customizing Default Scopes](#customizing-default-scopes).
### Add with CLI
```bash
claude mcp add --transport http \
--client-id YOUR_CLIENT_ID \
--client-secret \
--callback-port 8080 \
pipeshub PIPESHUB_INSTANCE_URL/mcp
```
> `--client-secret` without a value prompts for masked input. To skip the prompt, set the `MCP_CLIENT_SECRET` environment variable:
>
> ```bash
> MCP_CLIENT_SECRET=YOUR_CLIENT_SECRET claude mcp add --transport http \
> --client-id YOUR_CLIENT_ID \
> --client-secret \
> --callback-port 8080 \
> pipeshub PIPESHUB_INSTANCE_URL/mcp
> ```
To make it available across all projects:
```bash
claude mcp add --transport http --scope user \
--client-id YOUR_CLIENT_ID \
--client-secret \
--callback-port 8080 \
pipeshub PIPESHUB_INSTANCE_URL/mcp
```
### Add with JSON
```bash
claude mcp add-json pipeshub '{
"type": "http",
"url": "PIPESHUB_INSTANCE_URL/mcp",
"oauth": {
"clientId": "YOUR_CLIENT_ID",
"callbackPort": 8080
}
}' --client-secret
```
### Project-Scoped (`.mcp.json`)
Create a `.mcp.json` file in your project root. This can be committed to version control (secrets stay out via env vars):
```json
{
"mcpServers": {
"pipeshub": {
"type": "http",
"url": "${PIPESHUB_INSTANCE_URL}/mcp",
"oauth": {
"clientId": "${PIPESHUB_CLIENT_ID}",
"callbackPort": 8080
}
}
}
}
```
Set environment variables before launching Claude Code:
```bash
export PIPESHUB_INSTANCE_URL="https://app.pipeshub.com"
export PIPESHUB_CLIENT_ID="your-client-id"
```
> **Note:** The client secret is stored in the system keychain, not in config files. You'll be prompted to enter it when you first authenticate via `/mcp`.
### Authenticate
After adding the server, run `/mcp` inside Claude Code and follow the browser login flow. Tokens are stored securely and refreshed automatically.
### Verify
```bash
claude mcp list
claude mcp get pipeshub
```
</details>
<details>
<summary><strong>Gemini CLI</strong></summary>
Gemini CLI supports remote MCP servers with OAuth via `dynamic_discovery` (the default), which auto-discovers authorization and token endpoints from PipesHub's `/.well-known/oauth-protected-resource/mcp`.
### Option A: Settings File
Edit `~/.gemini/settings.json`:
```json
{
"mcpServers": {
"pipeshub": {
"url": "PIPESHUB_INSTANCE_URL/mcp",
"oauth": {
"clientId": "YOUR_CLIENT_ID",
"clientSecret": "YOUR_CLIENT_SECRET",
"scopes": [
"org:read", "org:write", "org:admin",
"user:read", "user:write", "user:invite", "user:delete",
"usergroup:read", "usergroup:write",
"team:read", "team:write",
"kb:read", "kb:write", "kb:delete", "kb:upload",
"semantic:read", "semantic:write", "semantic:delete",
"conversation:read", "conversation:write", "conversation:chat",
"agent:read", "agent:write", "agent:execute",
"connector:read", "connector:write", "connector:sync", "connector:delete",
"config:read", "config:write",
"document:read", "document:write", "document:delete",
"crawl:read", "crawl:write", "crawl:delete"
]
}
}
}
}
```
> **Note:** Adjust the `scopes` list to match what your OAuth app was granted. If you only need a subset of tools, you can limit the scopes accordingly.
### Option B: CLI Command
```bash
gemini mcp add --transport http pipeshub PIPESHUB_INSTANCE_URL/mcp
```
Then edit `~/.gemini/settings.json` to add the `oauth` block as shown above.
### Authenticate
Inside Gemini CLI, use the `/mcp auth` commands:
```bash
# List servers and their auth status
/mcp auth
# Authenticate with PipesHub (opens browser for login)
/mcp auth pipeshub
# Re-authenticate if tokens expire
/mcp auth pipeshub
```
On first connection, Gemini will automatically detect the 401 response, discover the OAuth endpoints, and open a browser for login. Tokens are stored securely in `~/.gemini/mcp-oauth-tokens.json` and refreshed automatically.
### Manage Servers
```bash
# List all configured servers
gemini mcp list
# Remove the server
gemini mcp remove pipeshub
# Temporarily disable/enable
gemini mcp disable pipeshub
gemini mcp enable pipeshub
```
### OAuth Configuration Properties
| Property | Required | Description |
|---|---|---|
| `clientId` | Yes | OAuth 2.0 Client ID from PipesHub |
| `clientSecret` | No | OAuth 2.0 Client Secret (for confidential clients) |
| `scopes` | No | OAuth scopes to request |
| `authorizationUrl` | No | Override authorization endpoint (auto-discovered by default) |
| `tokenUrl` | No | Override token endpoint (auto-discovered by default) |
| `redirectUri` | No | Override redirect URI (defaults to `http://localhost:7777/oauth/callback`) |
> **Note:** OAuth requires a local browser. It will not work in headless environments, remote SSH without X11 forwarding, or containers without browser access.
</details>
<details>
<summary><strong>Codex CLI</strong></summary>
Codex CLI ([OpenAI Codex](https://developers.openai.com/codex/mcp)) connects to remote MCP servers over **Streamable HTTP**, configured with a `[mcp_servers.<name>]` table in `~/.codex/config.toml` (or `.codex/config.toml` in your project root to scope it per-project). Codex's HTTP transport authenticates with a **bearer token** read from an environment variable, so pass a PipesHub JWT bearer token.
```toml
[mcp_servers.pipeshub]
url = "PIPESHUB_INSTANCE_URL/mcp"
bearer_token_env_var = "PIPESHUB_BEARER_TOKEN"
```
`bearer_token_env_var` is the **name** of the environment variable that holds the token — export it before launching Codex:
```bash
export PIPESHUB_BEARER_TOKEN="YOUR_BEARER_TOKEN"
```
> The token is the raw JWT, without the `Bearer` keyword.
Or add it with the CLI:
```bash
codex mcp add pipeshub \
--url PIPESHUB_INSTANCE_URL/mcp \
--bearer-token-env-var PIPESHUB_BEARER_TOKEN
```
> `--bearer-token-env-var` takes the **name of the environment variable** holding the token, not the token value itself.
### Verify
```bash
# List configured MCP servers
codex mcp list
# Inside the Codex TUI, view server status and available tools
/mcp
```
</details>
<details>
<summary><strong>Claude.ai (Web)</strong></summary>
Claude.ai supports custom connectors via remote MCP servers. This lets you use PipesHub tools directly in the Claude.ai web interface without any local setup.
> **Note:** This feature is currently in beta. Free plan users are limited to one custom connector.
### For Individual Users (Pro / Max Plans)
1. Go to [claude.ai](https://claude.ai) and navigate to **Settings > Connectors**
2. Click **Add custom connector** at the bottom of the Connectors section
3. Enter the MCP server URL:
```
PIPESHUB_INSTANCE_URL/mcp
```
4. Click **Advanced settings** and enter your OAuth credentials:
- **OAuth Client ID**: `YOUR_CLIENT_ID`
- **OAuth Client Secret**: `YOUR_CLIENT_SECRET`
5. Click **Add**
6. You'll be redirected to PipesHub's login page to authenticate and grant permissions
7. After authenticating, the connector will be active and PipesHub tools will be available in your Claude.ai conversations
### For Team / Enterprise Plans
**Organization Owners** must first add the connector:
1. Navigate to **Organization settings > Connectors**
2. Click **Add custom connector**
3. Enter the MCP server URL: `PIPESHUB_INSTANCE_URL/mcp`
4. Click **Advanced settings** and enter the OAuth Client ID and Client Secret
5. Click **Add**
**Team members** can then connect:
1. Go to **Settings > Connectors**
2. Find the PipesHub connector (marked with a "Custom" label)
3. Click **Connect** to authenticate via PipesHub's OAuth login
### Redirect URI
Claude.ai uses the following redirect URI for OAuth:
```
https://claude.ai/api/mcp/auth_callback
```
Register this as an allowed redirect URI in your PipesHub OAuth app.
### Security Notes
- Only connect to trusted MCP servers
- Review the permissions requested during the OAuth authentication flow
- Claude.ai interacts with PipesHub on your behalf using the granted OAuth token — your password is never shared
</details>
<details>
<summary><strong>LibreChat</strong></summary>
LibreChat supports remote MCP servers with OAuth authentication via its custom connectors UI. This lets you connect PipesHub tools to any model available in your LibreChat instance.
### Configuration
1. Log in to your LibreChat instance
2. Navigate to **MCP Servers** settings panel
3. Click **Add** to create a new custom MCP connector
4. Fill in the connector details:
- **Name**: `Pipeshub` (or any name you prefer)
- **MCP Server URL**: `PIPESHUB_INSTANCE_URL/mcp`
- **Transport**: Select **Streamable HTTPS**
- **Authentication**: Select **OAuth**
5. Enter your OAuth credentials:
- **Client ID**: `YOUR_CLIENT_ID`
- **Client Secret**: `YOUR_CLIENT_SECRET`
- **Authorization URL**: `PIPESHUB_INSTANCE_URL/api/v1/oauth2/authorize`
- **Token URL**: `PIPESHUB_INSTANCE_URL/api/v1/oauth2/token`
- **Scope**: `openid email` (or additional scopes as needed)
6. Check **I trust this application**
7. Click **Add** to save the connector
8. After adding, LibreChat will generate a **Redirect URI** displayed in the connector settings panel (next to the copy button). It follows this format:
```
http://localhost:3080/api/mcp/<server-identifier>/oauth/callback
```
9. **Copy the Redirect URI** and register it as an allowed redirect URI in your PipesHub OAuth app (see [Step 1](#step-1-create-an-oauth-app-in-pipeshub))
10. Return to the LibreChat connector and click **Update** to initiate the OAuth flow — you'll be redirected to PipesHub's login page to authenticate and grant permissions
### Redirect URI
LibreChat generates the redirect URI **after** the connector is created. The URI follows this format:
```
http://localhost:3080/api/mcp/<server-identifier>/oauth/callback
```
Where `<server-identifier>` is the unique identifier assigned by LibreChat (visible at the top of the connector settings as "Unique Server Identifier"). You must copy this URI and add it to your PipesHub OAuth app's allowed redirect URIs **before** authenticating.
> **Note:** If your LibreChat instance runs on a different host or port, the URI will reflect that (e.g., `https://chat.example.com/api/mcp/pipeshub/oauth/callback`).
### Scopes
LibreChat allows you to specify the OAuth scopes in the **Scope** field. Use a space-separated list:
```
openid email
```
To request PipesHub-specific scopes, add them to the scope field:
```
openid email org:read kb:read kb:write semantic:read conversation:read conversation:write conversation:chat agent:read agent:execute
```
> **Note:** The scopes you request must match the scopes granted to your OAuth app in PipesHub. See [Customizing Default Scopes](#customizing-default-scopes) for details.
</details>
---
## Local MCP Server (Stdio)
Instead of connecting to PipesHub's remote MCP endpoint, you can run the MCP server locally as a stdio process using the `@pipeshub-ai/mcp` npm package. This is useful when you prefer a local setup or need to work in environments where direct HTTP connections to the remote MCP endpoint aren't practical.
### Prerequisites
- Node.js 18+ installed
- A PipesHub instance URL
- A credential that acts as you, so results follow your permissions. Either:
- **No token in the config:** run `pipeshub login` once (`npm i -g @pipeshub-ai/mcp`; `--url <origin>` when self-hosted). With no `--bearer-auth`, `start` uses that sign-in, refreshes it (on PipesHub versions that issue device-login refresh tokens; otherwise it lasts an hour), and defaults `--server-url` to the instance you signed in to. It also reads `PIPESHUB_TOKEN` from the environment, sent only to `--server-url` or `PIPESHUB_BASE_URL`.
- **A [personal access token](https://docs.pipeshub.com/developer/personal-access-tokens)** (`phpat_…`), minted in PipesHub under Developer Settings and passed as `--bearer-auth`, as in the examples below.
With `pipeshub login`, a Claude Desktop entry is just:
```json
{
"mcpServers": {
"pipeshub": {
"command": "npx",
"args": ["@pipeshub-ai/mcp", "start"]
}
}
}
```
### Placeholders
Replace these in all configurations below:
| Placeholder | Description | Example |
|---|---|---|
| `PIPESHUB_INSTANCE_URL` | Your PipesHub instance URL | `https://app.pipeshub.com` |
| `YOUR_BEARER_TOKEN` | Your personal access token | `phpat_...` |
<details>
<summary><strong>Claude Desktop</strong></summary>
Configure in Claude Desktop settings (`claude_desktop_config.json`):
```json
{
"mcpServers": {
"pipeshub": {
"command": "npx",
"args": [
"@pipeshub-ai/mcp",
"start",
"--server-url",
"PIPESHUB_INSTANCE_URL",
"--bearer-auth",
"YOUR_BEARER_TOKEN"
]
}
}
}
```
</details>
<details>
<summary><strong>Cursor (Local)</strong></summary>
Open Cursor Settings > Tools and Integrations > New MCP Server, or edit your project's `.cursor/mcp.json`:
```json
{
"mcpServers": {
"pipeshub": {
"command": "npx",
"args": [
"@pipeshub-ai/mcp",
"start",
"--server-url",
"PIPESHUB_INSTANCE_URL",
"--bearer-auth",
"YOUR_BEARER_TOKEN"
]
}
}
}
```
</details>
<details>
<summary><strong>Claude Code CLI (Local)</strong></summary>
```bash
claude mcp add pipeshub -- npx -y @pipeshub-ai/mcp start \
--server-url PIPESHUB_INSTANCE_URL \
--bearer-auth YOUR_BEARER_TOKEN
```
</details>
<details>
<summary><strong>Gemini CLI (Local)</strong></summary>
```bash
gemini mcp add pipeshub -- npx -y @pipeshub-ai/mcp start \
--server-url PIPESHUB_INSTANCE_URL \
--bearer-auth YOUR_BEARER_TOKEN
```
</details>
<details>
<summary><strong>Codex CLI (Local)</strong></summary>
Run the MCP server as a local stdio process with your personal access token. Edit `~/.codex/config.toml` (or `.codex/config.toml` in your project root):
```toml
[mcp_servers.pipeshub]
command = "npx"
args = [
"-y",
"@pipeshub-ai/mcp",
"start",
"--server-url",
"PIPESHUB_INSTANCE_URL",
"--bearer-auth",
"YOUR_BEARER_TOKEN",
]
```
</details>
<details>
<summary><strong>VS Code</strong></summary>
Open Command Palette > `MCP: Open User Configuration`, then add:
```json
{
"mcpServers": {
"pipeshub": {
"command": "npx",
"args": [
"@pipeshub-ai/mcp",
"start",
"--server-url",
"PIPESHUB_INSTANCE_URL",
"--bearer-auth",
"YOUR_BEARER_TOKEN"
]
}
}
}
```
</details>
<details>
<summary><strong>Windsurf</strong></summary>
Open Windsurf Settings > Cascade > Manage MCPs > View raw config, then add:
```json
{
"mcpServers": {
"pipeshub": {
"command": "npx",
"args": [
"@pipeshub-ai/mcp",
"start",
"--server-url",
"PIPESHUB_INSTANCE_URL",
"--bearer-auth",
"YOUR_BEARER_TOKEN"
]
}
}
}
```
</details>
<details>
<summary><strong>Running from Source (Development)</strong></summary>
To run the local MCP server from a cloned repository instead of the npm package:
```bash
git clone https://github.com/pipeshub-ai/pipeshub-ai.git
cd pipeshub-ai
npm install
npm run build
node ./bin/mcp-server.js start --server-url PIPESHUB_INSTANCE_URL --bearer-auth YOUR_BEARER_TOKEN
```
For MCP client configuration, replace `npx @pipeshub-ai/mcp` with `node ./bin/mcp-server.js`:
```json
{
"mcpServers": {
"pipeshub": {
"command": "node",
"args": [
"./bin/mcp-server.js",
"start",
"--server-url",
"PIPESHUB_INSTANCE_URL",
"--bearer-auth",
"YOUR_BEARER_TOKEN"
]
}
}
}
```
To debug with MCP Inspector:
```bash
npx @modelcontextprotocol/inspector node ./bin/mcp-server.js start --server-url PIPESHUB_INSTANCE_URL --bearer-auth YOUR_BEARER_TOKEN
```
</details>
### CLI Help
For a full list of server arguments:
```bash
npx @pipeshub-ai/mcp --help
```
### Retries and Timeouts
The local server (`start` and `serve`) retries a PipesHub request that is refused with 429, 502, 503 or 504. It waits for as long as the `Retry-After` header asks, or, without one, backs off for half a second and then a second. It stops after three tries in all and passes the last answer on. A `Retry-After` of more than 30 seconds is not waited out: that answer is passed on at once, and the tool's error says how long PipesHub asked to wait. Only requests that are safe to repeat are retried (`GET`, `HEAD`, `OPTIONS`, `PUT`, `DELETE`). Search and chat are `POST`s and are never retried, because a repeat would be a second search or a second answer.
A request that gets no response within 60 seconds is abandoned. The limit covers the wait for PipesHub to start answering, not the whole answer, so a long chat response that is already streaming is not cut off.
Both are set with environment variables, in the MCP client's `env` block or with `--env NAME=VALUE`:
| Variable | Default | Meaning |
| --- | --- | --- |
| `PIPESHUB_MCP_MAX_ATTEMPTS` | `3` | Tries per request, the first one included, from 1 to 10. `1` turns retries off. |
| `PIPESHUB_MCP_TIMEOUT_MS` | `60000` | Milliseconds to wait for PipesHub to start answering. `0` turns the timeout off. |
---
## Read-only mode
Two tools write: `pipeshub_create_collection` and `pipeshub_upload` add collections and documents (they never delete or overwrite). They need the `kb:write` and `kb:upload` scopes. To serve read tools only from the local server, pass `--scope read` to `start` or `serve`. On the remote `/mcp`, a token without those scopes is not offered the write tools.
---
## Leaving `pipeshub_chat` out
The server offers all eleven tools, and its instructions and tool descriptions tell the agent to answer with `pipeshub_search` and `pipeshub_get_record_content` first. `pipeshub_chat` runs PipesHub's own assistant and is optional: for an agent that can call tools, it adds cost and time without adding accuracy. In PipesHub's evaluation (100 questions, Claude Opus as the agent), search and read alone found 88% of the facts at $0.54 and 78 seconds median per answer; adding chat also found 88%, at about $0.73 and 97 seconds. See [TOOLS.md](./TOOLS.md) for the full figures.
Nothing changes unless you choose to: by default every tool stays. To hold an agent to search and read, leave `pipeshub_chat` out, together with `pipeshub_agents`, which only picks an agent for chat:
- **Local stdio server:** set `PIPESHUB_MCP_CHAT=off` in the server's `env` block in your MCP client config, or pass `--env PIPESHUB_MCP_CHAT=off` to `start` or `serve`. The server then mounts nine tools, and its instructions no longer mention chat or agents. Unset or `on` keeps today's behaviour; any other value stops the server with an error that says what it accepts.
```json
"env": { "PIPESHUB_MCP_CHAT": "off" }
```
To pick tools one by one instead, list the ones to keep with `--tool` (for example `--tool pipeshub_search --tool pipeshub_get_record_content`). `--tool` does not change the instructions.
- **Remote endpoint (`/mcp`):** turn the tool off in the client. In Claude Code, run `claude --disallowedTools "mcp__pipeshub__pipeshub_chat,mcp__pipeshub__pipeshub_agents"`, where `pipeshub` is the name you gave the server, or put the same entries in the `permissions.deny` list of `.claude/settings.json`. In other clients, use the per-tool toggles in their MCP settings, where they offer them.
Keep `pipeshub_chat` when the client cannot run its own tool loop, when you want PipesHub's assistant to write the answer, to talk to a configured PipesHub agent, or to take an action through the user's connected tools.
---
## How It Works
### Architecture
```
AI Client (Cursor / Claude Code / Gemini CLI / Codex CLI / Claude.ai / LibreChat)
│
│ HTTP POST (JSON-RPC)
│ Authorization: Bearer <token>
▼
PIPESHUB_INSTANCE_URL/mcp
│
│ StreamableHTTP Transport
│ (stateless, per-request MCP server)
▼
PipesHub API (curated tool set — see TOOLS.md)
```
### OAuth Protected Resource Discovery
PipesHub exposes OAuth protected resource discovery at:
```
PIPESHUB_INSTANCE_URL/.well-known/oauth-protected-resource/mcp
```
This returns all OAuth endpoints automatically:
- Authorization: `PIPESHUB_INSTANCE_URL/api/v1/oauth2/authorize`
- Token: `PIPESHUB_INSTANCE_URL/api/v1/oauth2/token`
- Revocation: `PIPESHUB_INSTANCE_URL/api/v1/oauth2/revoke`
- JWKS: `PIPESHUB_INSTANCE_URL/.well-known/jwks.json`
---
## Troubleshooting
### "Incompatible auth server: does not support dynamic client registration"
This means the client is trying dynamic registration instead of using your pre-configured credentials. Make sure you passed `--client-id` and `--client-secret` (Claude Code) or the `auth` object (Cursor) correctly.
### Authentication fails / redirect error
- Ensure the **Redirect URI** in your OAuth app matches exactly what the client uses:
- **Cursor**: `cursor://anysphere.cursor-mcp/oauth/callback`
- **Claude Code**: `http://localhost:<callbackPort>/callback`
- **Claude.ai**: `https://claude.ai/api/mcp/auth_callback`
- **Gemini CLI**: `http://localhost:7777/oauth/callback`
- **LibreChat**: `http://localhost:3080/api/mcp/<server-identifier>/oauth/callback`
- Make sure the OAuth app is **active** (not suspended) in PipesHub
### Cannot reach MCP endpoint
- Verify the endpoint is accessible: `curl -X POST PIPESHUB_INSTANCE_URL/mcp` (should return 401, not connection error)
- Check that your PipesHub instance has MCP enabled
### Debugging with MCP Inspector
```bash
npx @modelcontextprotocol/inspector
```
Then connect to `PIPESHUB_INSTANCE_URL/mcp` with a Bearer token to test the endpoint directly.
---
## FAQ
<details>
<summary><strong>How do I update the scopes for my MCP integration?</strong></summary>
1. **Update the [`MCP_SCOPES`](https://github.com/pipeshub-ai/pipeshub-ai/blob/main/backend/env.template#L57) environment variable** on your PipesHub instance to include the new scopes you want exposed via the discovery endpoint.
2. **Update the OAuth app scopes** in PipesHub: go to **Settings > Developer Settings > OAuth Apps**, select your OAuth app, and add or remove scopes as needed.
3. **Re-authenticate the client** — existing tokens carry the old scopes, so you need to re-authenticate to get a new token with the updated scopes. For example:
- **Cursor**: Remove and re-add the MCP server, or clear the cached OAuth token and reconnect.
- **Claude Code**: Run `/mcp` and complete the browser login flow again.
- **Gemini CLI**: Run `/mcp auth pipeshub` to re-authenticate.
- **Codex CLI**: Update `PIPESHUB_BEARER_TOKEN` with a fresh token and restart Codex.
- **Claude.ai**: Disconnect and reconnect the connector in **Settings > Connectors**.
</details>