Kiro MCP Server Setup: Add and Verify Tool Connections

How to add MCP servers to Kiro, test each connection, and set up fenbs so Kiro and any other connected assistant start from the same project decisions and next steps.

Updated 5 min read

Kiro is an AI-powered IDE from AWS that supports MCP natively, letting you connect external tools and services to the assistant inside the IDE. You add MCP servers through a configuration file, and Kiro's assistant can then call those tools in any session. Adding fenbs as an MCP server means Kiro reads the same project decisions, rules, and Where we left off as every other assistant connected to your team — so the context is consistent regardless of which tool is open.

Kiro stores its MCP configuration in a mcp.json file in your project or user settings directory. The format is similar to the configuration used by Claude Desktop and other MCP clients: a JSON object with named server entries, each specifying a command, arguments, and environment variables.

Add an MCP server to Kiro

Open the MCP settings in Kiro — accessible through the IDE settings panel or by editing the mcp.json file directly. Each server entry needs a name, a command to start the server, and any arguments or environment variables it requires. Restart or reload the MCP configuration after editing and confirm the server appears in Kiro's tool panel.

Example mcp.json entry for a stdio server
{
  "mcpServers": {
    "my-server": {
      "command": "node",
      "args": ["/path/to/server/index.js"],
      "env": {
        "API_KEY": "your-key-here"
      }
    }
  }
}

For servers that require sensitive credentials, use environment variable references rather than hardcoding values in the configuration file. Kiro supports reading environment variables from the system environment or from a local .env file that is not committed to version control.

Verify each connection

After adding a server, verify that it works before using it for real work. Ask Kiro's assistant to list the available tools from the server and call one with a simple argument. Confirm the result is what you expected. If the server does not appear or the tool call fails, check the server logs — most servers write startup errors to stderr — and confirm the command and arguments are correct.

  • List the tools the server exposes and confirm the names and descriptions are correct.
  • Call a read-only tool with a valid argument and confirm the response.
  • Check that the server handles an invalid argument with an error rather than a crash.
  • Confirm the server is accessible without elevated privileges if it runs on a shared machine.

Connect fenbs to give Kiro consistent project context

fenbs connects to Kiro the same way as any other MCP server: add it to mcp.json with the fenbs MCP endpoint and sign in to approve access. Once connected, Kiro's assistant reads the same project rules, tasks, and Where we left off that Claude, Cursor, or any other connected tool reads. This matters when you use multiple tools or work in a team: decisions recorded once in fenbs reach every tool, rather than being repeated in each tool's project settings.

When Kiro starts a session and reads the fenbs context, it knows which tasks are in progress, what rules apply, and where the previous session ended. When the session closes, asking the assistant to save progress in fenbs means the next tool — or the next day's Kiro session — starts from the current state. See the Integrations page for the current setup guide.

Troubleshoot connection problems

If a server does not appear in Kiro after adding it to mcp.json, check three things: the command path is correct and the executable exists, the environment variables required by the server are set, and the configuration file is valid JSON. Kiro may show an error in its MCP settings panel, or the server's startup output may appear in Kiro's output panel. For stdio servers, the most common issue is a wrong path to the server script or a missing package that the server tries to import at startup.

Manage multiple MCP servers in Kiro

A Kiro project might use several MCP servers: one for source control, one for a database, one for project context. Keep the list short. Each server adds startup time, each tool call has latency, and a long list of tools makes it harder for the assistant to pick the right one. Add servers only when you have a specific, tested use case for each.

Name each server clearly in mcp.json so its purpose is obvious when you or a teammate revisits the configuration. A name like postgres-readonly is more useful than db when you are debugging why a tool call is failing three months later.

How to use Kiro MCP servers with fenbs

Connect fenbs to Kiro and record the project context there. Every assistant that connects to the same fenbs project reads the same starting point.

  1. Create a free account and start a project for the work you do in Kiro. Describe the goal and the current tasks.
  2. Write the first Decisions and rules: which operations the Kiro assistant may perform, which MCP servers it should use for specific tasks, and any constraints that always apply.
  3. Connect fenbs to Kiro via Integrations. Sign in to approve access and confirm fenbs tools appear in Kiro's tool panel.
  4. At the start of each Kiro session, ask the assistant to read the fenbs project context before starting work.
  5. Before closing Kiro, ask the assistant to save progress in fenbs so the next session — in Kiro or any other tool — starts from the current state.
How fenbs helps

Give Kiro and every other assistant the same project context

fenbs gives Kiro's assistant the same project decisions, rules, and Where we left off that every other connected tool reads. When you switch between Kiro, Claude, and other assistants, the context is consistent because it lives in fenbs, not in any one tool's settings. Create a free account and connect it to Kiro before your next session.

Create a free account

Free for one person · No credit card · Works with Claude, ChatGPT, Cursor and more

Questions people ask.

What is Kiro?

Kiro is an AI-powered IDE developed by AWS. It includes a built-in AI assistant and supports MCP, letting you connect external tools and services to the assistant directly from the IDE.

Where does Kiro store its MCP configuration?

Kiro uses an mcp.json configuration file, either per-project or in user settings. Check the Kiro documentation for the exact file location for your version, as it may vary between releases.

Can I use the same MCP servers in Kiro and Claude Desktop?

Yes. MCP servers are independent processes. You add the same server to each client's configuration. If the server uses stdio transport, each client spawns its own instance. If it uses the SSE transport, multiple clients can connect to the same running server.

Does fenbs work differently in Kiro than in other AI tools?

No. fenbs connects over MCP, which is the same standard regardless of the client. Kiro, Claude, Cursor, and other MCP-compatible clients all connect to fenbs the same way and get the same tools and context.

How do I stop a Kiro MCP server from running when I do not need it?

Remove or comment out the server entry in mcp.json and reload the configuration. Stdio servers are spawned on demand, so removing the configuration entry prevents Kiro from starting the server process.

More in Connect your AI tools

All Connect your AI tools articles

Start with one thing.

Free for one person · No credit card · Works with Claude, ChatGPT, Cursor and more

There is nothing to set up first. Write one line and you’ve started.