LaunchRepo with Kiro: where the browser config goes

Short answer

• For one project, save the entry in .kiro/settings/mcp.json; for all projects, in ~/.kiro/settings/mcp.json. The builder below writes either file, or a custom agent.
• Kiro uses the same "mcpServers" shape as Claude Code: "command" plus "args". If both files exist, Kiro merges them and the workspace wins.
• Leave approvals on for anything that types or clicks. If you like, let Kiro read pages without asking through autoApprove (in mcp.json) or allowedTools (in an agent).
• For a CAPTCHA or 2FA step, let Kiro stop, finish the check yourself in the same tab, then continue the conversation: in the CLI with kiro-cli chat --resume.
• Kiro needs a sign-in for everything, including its MCP commands. We checked what works before that and that every file from the builder starts the browser server.

LaunchRepo works with any coding agent that can read and write files in your workspace and control a browser. Kiro does both. This page documents a configuration and our checks; it is not a certification or partnership.

Build your mcp.json

Three choices, one file. Everything runs in your browser; nothing is sent anywhere.

{
  "mcpServers": {
    "chrome-devtools": {
      "command": "npx",
      "args": [
        "-y",
        "chrome-devtools-mcp@1.10.1",
        "--autoConnect",
        "--no-usage-statistics",
        "--no-performance-crux"
      ]
    }
  }
}

If a file already exists, add only the chrome-devtools entry inside its "mcpServers" object instead of replacing the file. Kiro picks up the change when you save.

1. IDE: open the command palette, run “Kiro: Open workspace MCP config (JSON)” or “Kiro: Open user MCP config (JSON)”, and paste the entry. MCP support must be enabled in Settings.
2. CLI: save the file at the path above; agents go in .kiro/agents/ and you start them with kiro-cli chat --agent launchrepo.
3. Ask Kiro to open https://example.com and read the page title. A correct title means the browser is connected.

Where Kiro looks for MCP servers

Three layers, highest first. Different server names add up; the same name is taken from the highest layer only.

1. 1
   Agent config

   The mcpServers field of a custom agent in .kiro/agents/ or ~/.kiro/agents/.

2. 2
   Workspace

   .kiro/settings/mcp.json in the project.

3. 3
   User

   ~/.kiro/settings/mcp.json in your home folder.

• An agent can switch a server off for itself by defining it with "disabled": true, even if the workspace file starts it.
• With "includeMcpJson": true, an agent also gets every server from both mcp.json files.
• KIRO_HOME moves the whole ~/.kiro folder, useful for a separate launch profile on the same machine.

Which Kiro reads what

Who approves browser actions

Kiro stacks several controls. For launch work, the rule of thumb is simple: reading may be automatic, typing and clicking should wait for you.

• Leave --trust-all-tools off whenever Kiro is connected to a signed-in Chrome.
• Workspace permission rules live in your home folder, not in the repository, so a cloned project cannot approve tools for you.
• The configuration reference shows plain tool names in autoApprove, which the builder uses. If Kiro still asks for a read-only tool, use the exact name Kiro shows in its approval prompt.

CAPTCHA, 2FA and resuming

When the form reaches a CAPTCHA, 2FA prompt or passkey, Kiro should stop and describe the one step you take. With approvals on, it cannot click past the check anyway.

Complete the check in the same Chrome tab. With your own Chrome (--autoConnect) the tab stays open even after the chat ends. The temporary Chrome (--isolated) closed together with the browser server in all 3 checks, so keep that chat open until you are done.

Then continue the same conversation. In the IDE it is the open chat. In the CLI, Kiro saves sessions per folder; resume the latest one, or pick one by ID.

Headless runs (kiro-cli chat --no-interactive) need a Kiro API key from a paid plan. For launch work with handoffs, the interactive session is the natural fit.

Resume the latest session in this folder

kiro-cli chat --resume

Resume a specific session

kiro-cli chat --resume-id <SESSION_ID>

List saved sessions

kiro-cli chat --list-sessions

How the handoff works, with a demo form →

What we checked without an account

Kiro needs a sign-in with GitHub, Google, AWS Builder ID or an organization before it runs any chat. We did not create an account. Instead we checked Kiro CLI 2.26.1 as far as it goes without one, 3 times in a fresh home folder, and started every file from the builder exactly as written.

Same server, three config dialects

Coming from another agent? The Chrome DevTools MCP entry carries over between Kiro and Claude Code almost unchanged. OpenCode needs a different shape.

Start LaunchRepo in Kiro

Open your LaunchRepo private workspace in Kiro, or start kiro-cli chat there, and paste the start prompt. The kit then leads onboarding, tests your Chrome connection on Example Domain and builds the run.

• The start prompt recommends Claude Code or Codex CLI for the best results. Kiro works with the same kit because it has the two things the kit needs: file access in the workspace and a browser through MCP.
• The agent option of the builder points Kiro at the kit’s AGENTS.md as its prompt, so the rules travel with the agent.
• CAPTCHA and 2FA steps come back to you as handoffs; you resume the session as shown above.

LaunchRepo start prompt

Recommend Claude Code CLI or Codex CLI in a terminal instead of the desktop apps for best results. If session metadata confirms desktop hosting, explicitly advise me to switch to the CLI and resume this same private workspace. Do not infer the host from a CLI engine version alone.

Read AGENTS.md, then follow "Start the dashboard for the owner". Create my private
workspace if I do not have one yet, start the dashboard, and give me the link.
Then lead ONBOARDING.md one phase at a time. Reuse recorded permissions instead of
asking again unless their scope is missing or conflicting. Actually test my visible
existing Chrome profile and intended identity as BROWSER-SETUP.md requires; use the
OS fallback if needed. Use one platform/browser worker by default. Build the normal
run only from suitable free-ready routes, keep research separate, and preserve any
draft that needs one exact human CAPTCHA/2FA action. Never claim browser control
bypasses or reduces those challenges.

• Connect your coding agent to your Chrome →
• Launch your SaaS on directories with Claude Code →
• LaunchRepo with OpenCode: browser MCP setup, tested →
• MCP server →

Not covered by our checks

• Loading the files inside a signed-in Kiro IDE or CLI, the approval prompts and the handoff itself. Those steps follow Kiro’s docs; we did not create a Kiro account.
• Which autoApprove spelling Kiro expects for MCP tools. The docs show both plain and prefixed names.
• A real AI model, --autoConnect to an everyday Chrome profile, Windows and Linux, Kiro Web and paid plans.

Sources

Every quote above is copied word for word from these pages, read on October 1, 2026. Tested behavior comes from our own runs described on this page.

FAQ

Where is the Kiro MCP config file?

In .kiro/settings/mcp.json inside your project for workspace servers, and in ~/.kiro/settings/mcp.json for servers you want everywhere. Custom agents can define servers too, in .kiro/agents/ or ~/.kiro/agents/.

What does the Kiro mcp.json for Chrome DevTools MCP look like?

An "mcpServers" object with a chrome-devtools entry: "command": "npx" and "args": ["-y", "chrome-devtools-mcp@1.10.1", "--autoConnect", "--no-usage-statistics", "--no-performance-crux"]. The builder on this page writes it for you.

Workspace or user config: which one wins in Kiro?

Kiro merges both. For a server with the same name, the workspace file wins, and an agent’s own mcpServers entry beats both.

Can Kiro use the browser without asking every time?

Yes, list tools under autoApprove in mcp.json or under allowedTools in an agent. For launch work, keep that to read-only tools such as take_snapshot and let typing and clicking ask.

How do I continue a Kiro session after a CAPTCHA?

Finish the check in the same Chrome tab, then return to the open chat in the IDE, or run kiro-cli chat --resume (or --resume-id <SESSION_ID>) in the same folder and ask Kiro to re-read the form before it submits.

Is Kiro officially supported by LaunchRepo?

LaunchRepo supports coding agents that have file and browser access; Kiro has both. This page documents a configuration and our checks without an account; it is not a certification or partnership, and the start prompt still recommends Claude Code or Codex CLI for the best results.