MCP Integration

Connect any Model Context Protocol-compatible server to extend jcode’s capabilities. MCP servers provide additional tools that merge seamlessly with the built-ins.

What Is MCP?

The Model Context Protocol (MCP) is an open protocol that allows AI models to safely interact with external tools and data sources. With MCP, jcode can:

Query databases
Search documentation
Manage cloud resources
Interact with APIs (GitHub, Linear, Notion, etc.)
Control browsers or other applications
And much more

jcode has built-in tools (file read/write, shell commands, search, etc.). Through MCP, you can add more tools without modifying jcode itself.

MCP Server Management

Add a Server

Via Command Line

# Add an HTTP/SSE server
jcode mcp add context7 https://mcp.context7.com/mcp

# Add with custom headers
jcode mcp add context7 https://mcp.context7.com/mcp \
  --header "CONTEXT7_API_KEY: your-key"

# Add a stdio server (local process)
jcode mcp add chrome-devtools -- npx chrome-devtools-mcp@latest

# Specify server type explicitly
jcode mcp add db -t http http://localhost:3001/mcp

Via Config File

Add servers to ~/.jcode/config.json:

{
  "mcp_servers": {
    "github": {
      "type": "stdio",
      "command": "gh-mcp"
    },
    "context7": {
      "type": "http",
      "url": "https://mcp.context7.com/mcp",
      "headers": {
        "CONTEXT7_API_KEY": "your-key"
      }
    },
    "remote-api": {
      "type": "sse",
      "url": "https://api.example.com/mcp",
      "headers": {
        "Authorization": "Bearer token..."
      }
    },
    "chrome-devtools": {
      "type": "stdio",
      "command": "npx",
      "args": ["chrome-devtools-mcp@latest"],
      "env": ["SOME_VAR=value"]
    }
  }
}

List Servers

jcode mcp list

Check Connectivity

jcode doctor

Server Types

Type	How It Works	Use Case
stdio	Runs a local subprocess, communicates via stdin/stdout	Local tools like `gh-mcp`, `npx` packages
http	Connects to an HTTP endpoint	Self-hosted MCP servers
sse	Server-Sent Events stream	Remote/cloud MCP servers

When using jcode mcp add, the server type is auto-detected from the URL. Use --type / -t to override.

Server Configuration

stdio Servers

Field	Description
`command`	Executable to run
`args`	Command arguments (array of strings)
`env`	Environment variables (`["KEY=VALUE"]`)

http / sse Servers

Field	Description
`url`	Server URL
`headers`	Custom HTTP headers (map of key-value pairs)

Loading Status

MCP servers initialize asynchronously after startup, so the TUI is usable immediately. The status bar shows live connection progress:

  Agent │ Model: openai / gpt-4o │ Approve: Ask │ [████░░░░░░] 2% │ MCP: 2/5

This means 2 out of 5 configured MCP servers are connected. Once all servers connect, the status updates automatically.

MCP servers auto-reconnect with exponential backoff if they disconnect. Check jcode doctor for detailed connection status.

Using MCP Tools

Once connected, MCP tools appear alongside built-in tools. The agent can use them just like any other tool:

  ⚙ Tool  mcp_github_search_issues  query="memory leak"

  ╭─────────────────────────────────────────────────────╮
  │  Found 3 issues:                                    │
  │  #142 - Memory leak in connection pool              │
  │  #98  - RSS grows over time                         │
  │  #45  - Large payloads not released                 │
  ╰─────────────────────────────────────────────────────╯

Security

MCP tools may access and operate external systems. Be aware of security implications.

Approval Mechanism

jcode requests user confirmation for sensitive operations (file modifications, command execution). MCP tools follow the same approval mechanism — all MCP tool calls prompt for confirmation by default.

Prompt Injection Risks

Content returned by MCP tools may contain malicious instructions attempting to trick the AI into performing dangerous operations. To stay safe:

Only use MCP servers from trusted sources
Review whether AI-proposed operations are reasonable
Keep manual approval enabled for high-risk operations

In unsafe mode (--unsafe), MCP tool operations are also auto-approved. Only use unsafe mode when you fully trust all configured MCP servers.

Quick Reference

Action	Command
Add a server	`jcode mcp add <name> <url-or-command>`
Add with headers	`jcode mcp add <name> <url> --header "Key: Value"`
Add stdio server	`jcode mcp add <name> -- <command> [args...]`
List servers	`jcode mcp list`
Check connectivity	`jcode doctor`
View in web UI	`GET /api/mcp`