Skip to main content
You can access a variety of tools in Agent Builder, including built-in tools and tools from remote MCP servers.

​
Built-in tools

Use these built-in tools to give your agents access to email, calendars, chat, project management, code hosting, spreadsheets/BI, search, social, and general web utilities.
Google, Slack, Linear, GitHub, and LinkedIn use OAuth. Exa, Tavily, Pylon, and Twitter/X use workspace secrets (API keys).

Gmail

Read and send email
  • Read emails (optionally include body, filter with search)
  • Send email or reply to an existing message
  • Create draft emails
  • Mark messages as read
  • Get a conversation thread
  • Apply or create labels
  • List mailbox labels

Google Calendar

Manage events
  • List events for a date
  • Get event details
  • Create new events

Google Sheets

Spreadsheets
  • Create spreadsheets
  • Read ranges

BigQuery

Analytics
  • Execute SQL queries

Slack

Send and read messages
  • Send a direct message to a user
  • Post a message to a channel
  • Reply in a thread
  • Read channel history
  • Read thread messages

LinkedIn

Post to profile
  • Publish a post with optional image or link

Twitter/X

  • Read a tweet by ID
  • Read recent posts from a list

GitHub

PRs, issues, and content
  • List pull requests
  • Get pull request details
  • Create issues and pull requests
  • Comment on issues and pull requests
  • Read repository files and list directories

Linear

Manage issues and teams
  • List teams and team members
  • List issues with filters
  • Get issue details
  • Create, update, or delete issues

Pylon

Issue management
  • List issues
  • Get issue details
  • Update issues

Search

  • Exa web search (optionally fetch page contents)
  • Exa LinkedIn profile search
  • Tavily web search

Web utilities

  • Read webpage text content
  • Extract image URLs and metadata
  • Notify user (for confirmations/updates)

​
Remote MCP server tools

An MCP (Model Context Protocol) server exposes tools that an agent can call at runtime. A remote MCP server:
  • Runs outside of LangSmith (usually over HTTPS).
  • Owns its own authentication and authorization.
  • Acts as a bridge between your agent and an external system.
LangSmith Agent Builder doesn’t execute these tools itself—it forwards requests to the MCP server and returns the results to the agent.

​
How it works

  • Agent Builder discovers tools from remote MCP servers via the standard MCP protocol.
  • Headers configured in your workspace are automatically attached when fetching tools or calling them. Headers are key-value pairs sent with every HTTP request to your MCP server. They’re commonly used for authentication (like API keys or bearer tokens), but can also provide configuration information, content types, or custom metadata.
  • Tools from remote servers are available alongside built-in tools in Agent Builder.
Runtime: Agent Builder automatically connects to your MCP server and uses its tools. The following sections show you how to connect a remote MCP server to Agent Builder:
  • General configuration: Step-by-step instructions for connecting any remote MCP server with authentication headers. Use this if you’re familiar with MCP servers and want a quick reference.
  • Example: Connecting a custom MCP server: A detailed walkthrough using a GitHub-based MCP server as an example. Use this if you want to see a complete end-to-end example with specific authentication details.

​
General configuration

Configure remote MCP servers in your LangSmith workspace:
1

Navigate to MCP Server settings

In the LangSmith UI, select Settings in the lower left corner, then choose MCP Servers from the left-hand navigation.
2

Add your MCP server

Click Add server and enter the URL of your remote MCP server (for example, https://mcp.example.com).
3

Configure authentication headers (if required)

If your MCP server requires authentication, add headers to authenticate requests. The most common pattern is using an Authorization bearer token:
  • Key: Authorization
  • Value: Bearer {{SECRET_NAME}}
How it works:
  • The {{SECRET_NAME}} placeholder references a workspace secret that stores your actual token value.
  • At runtime, Agent Builder automatically replaces {{SECRET_NAME}} with the secret’s value.
  • The header is attached to all requests to your MCP server.
Example: If you have a workspace secret named MCP_TOKEN with value sk_abc123xyz, configure:
  • Key: Authorization
  • Value: Bearer {{MCP_TOKEN}}
Agent Builder will send: Authorization: Bearer sk_abc123xyz
For instructions on adding workspace secrets, refer to How to add workspace secrets.
You can add multiple headers if your MCP server requires additional authentication or configuration parameters. Each header key-value pair is sent with every request to the server.
4

Save and verify

Save your configuration. Agent Builder will automatically discover available tools from your MCP server and make them available in your agents. The configured headers are applied to both tool discovery requests and tool execution requests.
Always use workspace secret placeholders (e.g., {{MCP_TOKEN}}) rather than hardcoding sensitive values directly in headers. Secrets are stored securely and resolved at runtime.

​
Example: Connecting a custom MCP server

Here’s a practical example of connecting Agent Builder to a GitHub MCP server that requires authentication:
1

Create a GitHub Authorization Token

The MCP server needs permission to access GitHub on your behalf. You’ll do this using a GitHub Personal Access Token (PAT).
  1. Go to GitHub → Settings → Developer settings.
  2. Open Personal access tokens.
  3. Create a Fine-grained token (recommended).
Grant read-only permissions:
  • Contents: Read
  • Issues: Read
  • Pull requests: Read
Once created, copy the token. You won’t be able to see it again.
Treat this token like a password. You can rotate or revoke it at any time without changing your agent.
2

Understand the Required Headers

Remote MCP servers typically authenticate using HTTP headers.For the GitHub MCP server, the required headers look like this:
  • Authorization proves who you are.
  • The MCP server validates the token.
  • Every tool call from the agent includes these headers.
Agent Builder will attach these headers automatically once configured.
3

Store the token as a workspace secret

In the LangSmith UI:
  1. Navigate to Settings > Workspaces > Secrets.
  2. Click Add secret.
  3. Name: GITHUB_TOKEN (or any descriptive name).
  4. Value: Your authentication token.
  5. Save the secret.
4

Configure the MCP server

In Settings > MCP Servers:
  1. Click Add server.
  2. Add a Name for the MCP server.
  3. URL: Enter your MCP server URL (e.g., https://mcp-github.example.com)
  4. Add authentication header:
    • Key: Authorization
    • Value: Bearer {{GITHUB_TOKEN}}
  5. Save the configuration.
5

Use tools in your agent

The tools from your MCP server are now available in Agent Builder. When you create or edit an agent, you’ll see these tools alongside the built-in tools. All requests to your MCP server will include the authentication header automatically.Once connected, the workflow for this example looks like this:
  1. The agent decides it needs GitHub data.
  2. It selects a tool exposed by the MCP server.
  3. LangSmith forwards the request to the remote MCP server.
  4. The server authenticates using your token.
  5. GitHub data is fetched and returned.
  6. The agent receives structured results and continues reasoning.
Different MCP servers may use different authentication methods:
  • Authorization: Bearer {{TOKEN}} (most common)
  • X-API-Key: {{API_KEY}}
  • Custom headers specific to your implementation
  • Multiple headers for authentication
Always check your MCP server’s documentation for the correct authentication header format.
Edit this page on GitHub or file an issue.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.