Using MCP Servers with Warp
# Using MCP Servers with Warp import VideoEmbed from '@components/VideoEmbed.astro'; <VideoEmbed url="https://www.youtube.com/watch?v=8vn2brhJrF8" /> ### 1. What Are MCP Servers? MCP (Model Context Protocol) servers let Warp agents connect to external systems like GitHub, Linear, or Jira — so they can read, write, and reason about those systems natively. Each MCP server adds its own “tools” to Warp’s agent.\ For example: * The Linear MCP Server handles tickets. * The GitHub MCP Server handles pull requests and issues. --- ### 2. Problem Setup Andrew starts with Warp’s universal input, but it doesn’t yet know what a “ticket” is. > Prompt Example:\ > “Help me solve this ticket.” At this point, Warp can’t find or interpret the ticket, because no MCP server is connected. --- ### 3. Adding the Linear MCP Server To connect Linear: 1. Open the **MCP Panel** in Warp 2. Click **Add Server** 3. Paste in the JSON configuration for the Linear MCP Server Once added, Warp: * Starts the MCP server * Loads its tools (e.g., `get_ticket`, `update_ticket`, `create_ticket`) * Makes them available to the agent instantly --- ### 4. Using Rules with MCP Servers Andrew adds a rule called **check-linear**, which helps the agent automatically associate “tickets” with the Linear MCP Server. > **Rule Example:**\ > “When the user says ‘ticket,’ check Linear.” Rules make context switching between systems seamless — the agent doesn’t need reminders. --- ### 5. Dynamic Context Loading Warp’s MCP support is **dynamic**: * You can start a conversation without any connected MCPs * Add one mid-session * The agent automatically updates its context on the next message No need to restart Warp or reset your session. --- ### 6. Running the Task After adding Linear, Andrew runs: > “Help me solve this ticket.” Now the agent: * Queries Linear for the ticket * Pulls all related context * Reads the codebase for linked references * Generates the appropriate fix He verifies the output by running: ```bash cargo run ```Connect MCP servers to Warp's agent, add Rules for automatic tool selection, and resolve tickets using external systems like Linear.
1. What Are MCP Servers?
Section titled “1. What Are MCP Servers?”MCP (Model Context Protocol) servers let Warp agents connect to external systems like GitHub, Linear, or Jira — so they can read, write, and reason about those systems natively.
Each MCP server adds its own “tools” to Warp’s agent.
For example:
- The Linear MCP Server handles tickets.
- The GitHub MCP Server handles pull requests and issues.
2. Problem Setup
Section titled “2. Problem Setup”Andrew starts with Warp’s universal input, but it doesn’t yet know what a “ticket” is.
Prompt Example:
“Help me solve this ticket.”
At this point, Warp can’t find or interpret the ticket, because no MCP server is connected.
3. Adding the Linear MCP Server
Section titled “3. Adding the Linear MCP Server”To connect Linear:
- Open the MCP Panel in Warp
- Click Add Server
- Paste in the JSON configuration for the Linear MCP Server
Once added, Warp:
- Starts the MCP server
- Loads its tools (e.g.,
get_ticket,update_ticket,create_ticket) - Makes them available to the agent instantly
4. Using Rules with MCP Servers
Section titled “4. Using Rules with MCP Servers”Andrew adds a rule called check-linear, which helps the agent automatically associate “tickets” with the Linear MCP Server.
Rule Example:
“When the user says ‘ticket,’ check Linear.”
Rules make context switching between systems seamless — the agent doesn’t need reminders.
5. Dynamic Context Loading
Section titled “5. Dynamic Context Loading”Warp’s MCP support is dynamic:
- You can start a conversation without any connected MCPs
- Add one mid-session
- The agent automatically updates its context on the next message
No need to restart Warp or reset your session.
6. Running the Task
Section titled “6. Running the Task”After adding Linear, Andrew runs:
“Help me solve this ticket.”
Now the agent:
- Queries Linear for the ticket
- Pulls all related context
- Reads the codebase for linked references
- Generates the appropriate fix
He verifies the output by running:
cargo run