code

Frequently Asked Questions

1. How is Kombai different from general-purpose agents?
   Kombai applies stack-specific best practices for 400+ frontend libraries, reuses components from your repo through Context Graphs, parses Figma files natively, verifies its work in a real browser, and ships with frontend-specific modes for planning and debugging.

2. Which frontend frameworks and libraries does Kombai support?
   Kombai applies human-tested best practices for over 400 frontend libraries covering React, Vue, Svelte, Angular, Next.js, Tailwind, shadcn/ui, MUI, Chakra, Storybook, and most popular charting, form, animation, and state-management libraries. It scans your repo first, figures out which ones are actually in use, and adapts its conventions to match.

3. Do I need a Figma plugin or MCP server to work with designs?
   No. Kombai's Figma integration is native, so you can hand it a Figma file directly. It parses complex files, fixes common issues like messy grouping or invisible layers, and pulls out icons and assets without needing any additional plugin or MCP server in the loop.

4. How does Kombai reuse code from my repo, npm packages, and Storybook?
   Through Context Graphs, a semantic index of your repo, your dependencies, and your Storybook. Instead of generating duplicate components or one-off styles, Kombai looks up what already exists (components, hooks, design tokens, types, constants) and reuses it, so your codebase stays consistent rather than fragmenting over time.

5. Which AI models does Kombai use, and how does it choose?
   Kombai uses Agent Variants. It continuously benchmarks top frontier models on common frontend tasks and routes each request to the model that performs best for the work, within the cost tier you've selected. You don't have to pick a model per task; Kombai handles routing automatically.

6. Can I run multiple tasks in parallel?
   Yes. You can spin up parallel agent threads to work on independent tasks simultaneously, instead of waiting for one to finish before starting the next. Useful when you want one agent reviewing while another is implementing.

7. What happens on long sessions? Will I hit context limits?
   Kombai has auto-compaction that intelligently summarizes long threads as they grow, so you don't run into context ceilings. You can disable it or trigger compaction manually whenever you want more control.

8. Can I extend Kombai with custom commands, skills, or MCP servers?
   Yes. You can define custom slash commands and add your own skills for workflows specific to your team, alongside the built-in Skills Library covering wireframing, architectural diagrams, SEO audits, and accessibility checks. MCP is fully supported for connecting external services, tools, and data sources, and Kombai also has built-in web fetch and terminal access.