Both Tandem Browser and Dia (from The Browser Company) take AI seriously as part of the browsing experience. They differ on openness, model choice, and design philosophy — Dia leads with refined consumer design; Tandem Browser leads with developer-facing flexibility. This page lays out the honest tradeoffs.
| Tandem Browser | Dia | |
|---|---|---|
| License | MIT (open source) | Proprietary |
| Source available | Yes — full repo on GitHub | No |
| Where it runs | Your machine, local-first | Cloud-backed |
| Model choice | Bring your own (Claude, GPT, Gemini, OpenClaw, Ollama, LM Studio, custom) | Anthropic-managed |
| Works offline | Yes (with local model) | No — needs cloud |
| MCP support | Native — 250+ tools | No |
| Multi-agent | Yes — multiple agents in one browser | No |
| Remote agents over Tailscale | Yes | N/A |
| UX polish | Developer preview — rough edges | Production-grade, design-led |
| Install effort | Clone repo, npm install | Download installer |
| Support | Community via GitHub Discussions | Paid product with support |
| Price | Free | Subscription tiers |
Both Tandem Browser and Dia recognize that browsers built before AI weren't designed for this kind of collaboration. Dia's polished design-led approach and Tandem Browser's open-source developer-facing approach answer the same question — "how should browsers and AI work together" — from genuinely different starting points.