Running Chrome DevTools MCP in Multiple Claude Code Sessions

7.3 relevance

Guide on running Chrome DevTools MCP with Claude Code; directly relevant to AI agent orchestration.

AI/ML dev.to

Answer the board on AI readiness

Arrive with evidence, not opinion. See how your foundation benchmarks against your peers, get a clear read, and leave with a ranked place to start.

Why It Matters

For engineers running AI agent pipelines across multiple Claude Code terminals, this solves a silent collision bug and provides an architectural insight for multi-client browser automation patterns.

Key Takeaways

Replace the default npx invocation with a wrapper that assigns per-terminal profiles via git-root-hash/fallback-to-tty, and flip to shared-browser mode if agents need persistent login sessions.

Running Chrome DevTools MCP in Multiple Claude Code Sessions

Summary

A lock on Chrome's user-data-dir prevents multiple Claude Code sessions from sharing a single chrome-devtools-mcp instance. A custom wrapper script resolves the conflict by assigning each terminal a unique, persistent browser profile through a hash of the Git repo root and a fallback that appends the TTY name. An unexpected finding after a week of subagent testing was that partitioned profiles break authentication-dependent workflows, making a shared browser with multiple MCP clients the preferable default.

Author

Dale Nguyen