InkPal is live at inkpal.ai

Six months ago I started building a tool because the AI assistants I worked with every day didn't understand Flutter. They wrote generic code. They hallucinated package APIs. They missed Flutter-specific gotchas — render-object lifecycles, widget tree depth, platform channel pitfalls.

Today InkPal is live at inkpal.ai. It's one MCP server with 179 Flutter-specific tools across 12 domains. Plug it into Claude, Cursor, Windsurf, or Gemini, and your AI assistant goes from "knows about Flutter" to "knows YOUR Flutter codebase."

What you get

1. Pattern search across 14,000+ curated Flutter patterns

Semantic search over real production patterns — state management, navigation, data fetching, animation, error handling. Indexed in Postgres pgvector. Your AI gets context, not Stack Overflow guesses.

2. Figma → Flutter pipeline that respects your project

Import a Figma design, generate widgets that match your project's design tokens, state management pattern, and navigation library. Not boilerplate Material — your codebase's idiom.

3. Visual regression in one tool call

Capture baselines for every route. Pixel-diff after every change with pixelmatch. Catches the 1px shifts your eyes miss. Three-line setup; HTML report with side-by-side diffs.

4. Runtime control via inkpal_bridge

Our Dart package on pub.dev — 33+ VM service extensions, 41 WebSocket commands, zero third-party deps. Your AI can tap, scroll, screenshot, hot-reload, evaluate Dart in your running app. Three modes: free (UI inspect), pro (interaction + telemetry), studio (network mock + multi-device).

5. Safety gates before any edit lands

Banned-package check. Secret scanner. Blast-radius analysis (which files import this?). Naming convention enforcer. Eighteen gates total. Generated code never lands without passing them.

6. Production monitoring hooks

Connect Sentry / Crashlytics / your analytics. Get crash reports + user flows + funnels surfaced as MCP tool results. AI roots-causes a stack trace in seconds.

How it actually feels

The shift from a generic AI to one running InkPal is the same shift as going from a senior dev who's never seen your codebase to one who's been on the team for six months. They speak your patterns. They cite your widgets. They don't introduce a new state-management library when you're already using Riverpod.

What's free vs paid

Free tier — no card: Pattern search (100/day), error lookup, Figma imports (5/month), bridge UI inspection.

Pro: All free tools at higher limits + bridge interaction + visual regression + production monitoring.

Studio: Everything + multi-device + network mocking + team DNA sharing.

Pricing details: inkpal.ai/pricing.

What's next

This week:

More IDE integrations (we currently support Claude Code, Cursor, Windsurf, Gemini CLI, generic MCP)
An Autopilot layer — background agents that watch Sentry, run nightly visual regression, ship dependency updates
Showcase apps — full open-source Flutter apps built end-to-end with InkPal so you can see the workflow

This month:

Custom auth domain so the Sign-in screens read "inkpal.ai" instead of the supabase project ID
More language coverage in the pattern index (currently English-only docstrings)
A /status page with public uptime data

Try it

# Free tier — sign up at inkpal.ai for a key
npx inkpal install --key <your-key>
 
# Then add to your Flutter app for runtime tools
flutter pub add inkpal_bridge

Setup guide for every IDE: inkpal.ai/setup.

Thank you

To the Flutter community, my early testers, and everyone who told me "it'd be useful if it could…" — this exists because of you. If you build with InkPal and find a gap, tell me. The roadmap is the gaps you find.

— Tejpal