Understand the code you didn't write.
Parse points at a GitHub repo and turns it into a guided course — lessons, quizzes, hands-on practice, and a debugging assistant. Ramp on an unfamiliar codebase in hours, not weeks.
- Lessons built from the real files in your repo
- Auto-generated quizzes and graded hands-on practice
- Stack-trace debug assistant maps errors back to the code
- Stays in sync — re-analyzes on every push
Affixly Parse is a code-comprehension tool that turns any GitHub repository into a guided course. It reads the repo directly, organizes the code into themed sections, and generates lessons, multiple-choice quizzes, and hands-on challenges from the actual files — plus a stack-trace debugging assistant that maps errors back to the exact function. Built for anyone ramping on a codebase they didn’t write.
A repo someone else wrote, and no map.
Joining a team, inheriting a legacy project, or onboarding onto open source means staring at thousands of lines someone else wrote, with no map. Reading top to bottom doesn't work and the docs are stale. Parse builds the map and teaches you the code automatically.
Connect a repo. Parse turns it into a course.
Connect a repo
Sign in with GitHub and pick a repository. Parse reads it directly; your access token is encrypted at rest.
Parse analyzes it
Detects the stack and primary language, then organizes the code into themed sections — Auth, Database, Routing, State, Infrastructure, Tests, plus a catch-all.
Learn section by section
Each unit pairs a real snippet from your repo with a plain-English explanation and a real-world analogy.
Test yourself
Auto-generated multiple-choice quizzes drawn from the actual code, not generic theory.
Practice hands-on
Open-ended challenges grounded in real files, with starter code and progressive hints. Submit and get graded against the original.
Track progress
Per-section read status, quiz scores, completed practice, and a per-repo summary.
Everything keyed to your actual repo.
Built from your actual code
Every lesson, quiz, and challenge references the real files in the repo you connected. Not generic tutorials.
Stack-trace debug assistant
Paste an error and Parse maps it back to the exact file and function, explains what went wrong, and links the relevant lesson.
Language cheat-sheets
Instant reference for the repo’s primary language — Python, JavaScript, TypeScript, Go — with syntax tables and inline keyword tooltips.
Search across the codebase
Find anything by title, file, or explanation.
Stays in sync
Connect a webhook and Parse re-analyzes incrementally on every push to the default branch. Only changed files regenerate, so progress and existing lessons are preserved.
Real-world analogies on every unit
Turns "I read it" into "I get it."
For anyone dropped into unfamiliar code.
Connecting a repo requires trust.
Parse reads your code to teach you — nothing more.
Access tokens encrypted at rest
Your GitHub access token is encrypted before storage and never kept in plaintext.
Read-only by design
Parse reads your code to teach you. It does not modify your repository.
At launch: use your own AI key and file limits are uncapped.
At launch, you can connect your own Anthropic, OpenAI, or Gemini key in Settings. Parse will route analysis and AI actions through your key — and files analyzed and AI actions stop counting against your plan.
Works with Anthropic, OpenAI, or Gemini
Connect any one (or all three). Parse routes each AI task to the right provider for you.
Usage on your key doesn't count toward AI limits
Files analyzed and AI actions only count against your plan when Parse uses Affixly's keys. On your key they run uncapped.
Keys are encrypted and never leave the server
Stored encrypted at rest, used only server-side. Never bundled, never logged.
Results may differ from the default models
Parse is calibrated on Claude Opus. Running on other providers or models may change explanation quality.
At launch: query Parse from your AI assistant.
At launch, you can add Parse as a connector in Claude or Claude Code using an API key from your account settings. Any MCP client will work.
“Explain this stack trace against my repo.”
“What files changed in the last push and should I re-read any lessons?”
“Show me my quiz scores across all sections.”
“What does the Auth module in this repo actually do?”
At launch: available in Claude Code and any MCP client with an API key. Entering API keys stays web-only for security.
Priced on repos and files, not seats.
Parse is in pre-launch. Waitlist members get founding rates. Team pricing is per seat with a minimum of three seats.
Free
$0/mo
1 public repo · 150 files · 10 AI actions per month
- Lessons, quizzes, and hands-on practice
- Stack-trace debug assistant
- Language cheat-sheets
- Public repos only
Pro
$19/mo
5 repos · 1,000 files · 100 AI actions per month
- Everything in Free
- Private repos
- Bring your own AI key — file and action limits uncapped on your key
- Work with it from Claude (MCP)
Team
$19/seat/mo
10 repos · 1,000 files · 50 AI actions per seat, pooled
- Everything in Pro
- Shared team repos
- Per-member GitHub access control
- Bring your own AI key — limits uncapped on your key
- Work with it from Claude (MCP)
Minimum 3 seats.
One file = one code file analyzed from your connected repo. Unchanged files aren’t re-analyzed. Terms · Privacy
Pricing questions
What is a file in Parse?
One file is one code file analyzed from your connected repository. Files whose content hasn't changed since the last analysis aren't re-analyzed and don't consume from your monthly allowance.
What is an AI action in Parse?
AI actions are debug explanations, scenario generation, scenario evaluation, and code explanations. Viewing existing lessons and quizzes doesn't count.
What does "Bring your own AI key" do in Parse?
With your own Anthropic, OpenAI, or Gemini key, files analyzed and AI actions no longer count against your plan limits — they run on your key instead. Non-AI limits (repos, seats) stay. Keys are encrypted at rest and never leave the server. Results may differ slightly from the default models.
What happens when I reach my plan limit?
Processing pauses — there is no surprise bill. Your existing content stays accessible. To upgrade, email hello@affixly.ai with your account and we will sort it out. Self-serve billing is coming in a future update.
What is MCP?
MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude talk to external tools. On the products that support it, you add the product as a connector in Claude or Claude Code using an API key from your account settings, and you can then query it directly from your AI assistant.
Be first to ramp with Parse.
Parse is coming soon. Join the waitlist to get early access when we open the doors.
Stop reading code top to bottom.
Parse turns the repo you were just handed into a course built from its own files. Join the waitlist for early access.