Open source · Free forever

AI tools that
learn with you

Machina is a suite of developer tools powered by AI — that actually improve over time by tracking your workflow patterns, errors, and decisions session by session.

★ View on GitHub See all tools →

MIT licensed · runs locally · no cloud required

See it in action

Watch the tools work

LearnBoard
Your AI finally remembers you. No re-explaining every session.
BugCapture
47 seconds of recording. Bug fixed in under 2 minutes.
PromptBoard
Build AI prompts visually. Canvas, voice, images — one .md export.
Stay in the loop

New tools, lessons, and AI workflow insights

No spam. Monthly digest of what was built, what was learned, and what's coming next.

No spam · Unsubscribe anytime · Built with Brevo

The core innovation
Your AI's memory,
made visible and editable

LearnBoard is the interface to the layer that makes Machina unique. Every lesson the AI learns, every pattern it recognizes, every workflow preference — stored in a structured file and surfaced here as a living dashboard you can read, search, and edit in real time.

1
See what the AI has learned — lessons, patterns, observations, vote history — all in searchable tables.
2
Edit and correct — click any cell to update a lesson, add a new row, or remove outdated knowledge.
3
Saved automatically — changes write back to the source file with a timestamped backup before every save.
Try LearnBoard → Install & run locally →
LearnBoard — AI Learning System
23
Lessons
25
Tools
32
Suggestions
50%
Shipped
#Lesson learned
15 Always prefer local/free solutions — never propose paid APIs without exhausting local alternatives first.
18 User prefers autonomous tools that find context on their own — not manual forms to fill in.
23 Always restart via the bash script that rebuilds nvm — nohup node fails silently without nvm env.
2 unsaved changes · backup enabled Save to CLAUDE.md
Featured tool
BugCapture — Turn bugs into
AI-ready context

Record your screen and audio while reproducing a bug. BugCapture transcribes your voice with offline Whisper, extracts sequential screenshots, and packages everything into a single .md file your AI agent can understand and act on immediately.

1
Record your screen + voice while reproducing the bug — no upload, fully local.
2
Export a structured .md with base64 screenshots + Whisper transcription.
3
Drop it into Copilot Agent or Claude — the AI sees the bug, reads your description, finds the fix.
Try BugCapture → Install & run locally →
bugcapture_report.md
# Bug Report
Duration: 00:34  Whisper ✓
Frames: 11 screenshots @ 1/3s
## Transcript
"The form fields aren't aligning after the Fabrik update — the left column overlaps the label on mobile..."
## Screenshot [3/11]
![frame_003](data:image/png;base64,iVBOR...)
→ Copilot found root cause in 40s
Visual prompt builder
PromptBoard — Build AI
prompts visually

Complex tasks don't fit in a text box. PromptBoard gives you a drag-and-drop canvas: add text blocks, drop screenshots, draw flow diagrams, connect them with labeled arrows. Export everything as a structured .md file any AI can parse immediately. No server, no install — just open the file.

1
Build on canvas — drag text, image, and flow blocks. Speak notes directly into any block via voice dictation.
2
Connect with arrows — label every relationship. The AI reads structure, not a wall of text.
3
Export as .md — one structured brief with embedded screenshots. Paste into Claude, ChatGPT, or Copilot.
Try PromptBoard → Install & run locally →
promptboard_brief.md
# Fix the checkout form
**[GOAL]** Cart component won't submit after refactor
**[CONSTRAINTS]** No new deps · TypeScript strict · <50 lines
**[SCREENSHOT]**
![cart-error](data:image/png;base64,iVBOR...)
## Flow
Cart validates → Payment API? POST /checkout
Payment API? → Show error state
→ Claude found root cause in 23s
Work with us
Need AI in your
development workflow?

We help companies integrate AI agents into their existing dev processes — without disrupting what already works. From custom tool development to training your team on AI-augmented debugging workflows.

Custom AI tooling for your stack
AI-augmented debugging workflow setup
Team training on AI-assisted development
Legacy codebase AI onboarding
Get in touch →
~40s
avg time for AI to
find a bug with BugCapture
100%
local processing
no data leaves your machine
How it works

The learning loop

Every session feeds the next. The AI accumulates context, preferences and lessons — so it gets more useful over time without any manual configuration.

01
You work with the tools
Build, debug, analyze — the tools run locally, capture context, and generate AI-ready outputs for Copilot or Claude.
02
The AI records what it learns
After significant sessions, the learning layer is updated: new lessons, workflow patterns, corrected mistakes, validated approaches.
03
Next session starts smarter
The AI reads the learning layer at session start. It already knows your preferences, past mistakes, and the context of your projects.