Connect a GitHub repository. CodeLens studies the architecture, the database, the auth flow and the git history before you ever ask a question — then answers grounded in how your project actually works.
Understanding comes before the chat.
Connect GitHub
Point CodeLens at any repository, public or private.
Clone & index
The repo is cloned and every file is indexed.
Detect the stack
Framework, dependencies, database, APIs, auth — all detected automatically.
Build the code graph
Relationships between files and modules are mapped, not just listed.
Read the history & docs
Git history, README and .env.example are folded into the analysis.
Generate the Project Brain
A structured, persistent model of how your project actually works.
Ask anything
Chat with your codebase, grounded in what it actually contains.
Teams that stopped re-explaining their codebase
CodeLens mapped our entire auth flow in minutes. That used to take a new backend hire the better part of a week.
We stopped pasting entire files into a chat window. CodeLens already knows how our checkout flow connects to Stripe and the webhook handler.
CodeLens flagged a missing idempotency check on our webhook before we'd even asked it a question. That one would've cost us a very bad night.
CodeLens mapped our entire auth flow in minutes. That used to take a new backend hire the better part of a week.
We stopped pasting entire files into a chat window. CodeLens already knows how our checkout flow connects to Stripe and the webhook handler.
CodeLens flagged a missing idempotency check on our webhook before we'd even asked it a question. That one would've cost us a very bad night.
Every push, CodeLens re-analyzes just what changed. Three months in and it still understands our codebase better than most of our contractors do.
We connected the repo, asked CodeLens how our subscription logic worked, and got an answer that pointed straight at the right five files. A generic chatbot never does that.
Ask CodeLens what depends on Redis and it traces the real call paths, not a guess based on package.json.
Every push, CodeLens re-analyzes just what changed. Three months in and it still understands our codebase better than most of our contractors do.
We connected the repo, asked CodeLens how our subscription logic worked, and got an answer that pointed straight at the right five files. A generic chatbot never does that.
Ask CodeLens what depends on Redis and it traces the real call paths, not a guess based on package.json.
Onboarding used to take a new hire two weeks before they could ship anything real. With CodeLens it's closer to two days.
We pointed CodeLens at three repos and had a full picture of how our services actually talk to each other within minutes, not a full sprint of reading code.
CodeLens's git history analysis surfaced context our own README had quietly gone stale on months ago.
Onboarding used to take a new hire two weeks before they could ship anything real. With CodeLens it's closer to two days.
We pointed CodeLens at three repos and had a full picture of how our services actually talk to each other within minutes, not a full sprint of reading code.
CodeLens's git history analysis surfaced context our own README had quietly gone stale on months ago.
What CodeLens actually does
Understands before it answers
CodeLens maps architecture, auth, payments, database and API flows before the first question — not a generic search over your files.
Stays current automatically
Every push, only what changed gets re-analyzed. The Project Brain still understands your project months later.
Grounded, not generic
Ask how the payment flow works or what breaks if you add an affiliate system, and get answers traced through your real files.
Works with your existing tools
Export the Project Brain to Claude Code, Cursor or any MCP-compatible agent. CodeLens is the context layer, not a replacement.
Ask with the model you trust
Every model reads the same Project Brain. Pick the one that fits the question — from the deepest reasoning to the fastest answer.
- Most capable
Claude Opus 5
Anthropic
The deepest reasoning in the lineup. Best for tracing a change across dozens of files or reviewing how an entire subsystem fits together.
Deepest reasoningMulti-file changes GPT-5.5
OpenAI
A strong all-rounder with broad tool support — a dependable pick for everyday debugging and code questions.
Strong reasoningBroad toolingGemini 3.1 Pro
Google
A massive 1M-token context window, built for questions that span an entire monorepo at once.
Massive contextMultimodalClaude Sonnet 5
Anthropic
The fast default for back-and-forth questions while you work — grounded answers in a couple of seconds.
Fast responsesGreat defaultLlama 4 Maverick
Meta
Open-weight and self-hostable, for teams that need the Project Brain to stay inside their own infrastructure.
Open weightsSelf-hostable- Fast & efficient
DeepSeek V4 Flash
DeepSeek
The quickest option in the lineup — built for short, frequent questions without the wait, and still grounded in your real code. Included on the free plan.
FastestMost efficient
Pricing that scales with your project
Starter
- Up to 3 repositories
- 25,000 tokens/day
- Full Project Brain
- Auto re-analysis on push
Pro
- Up to 10 repositories
- 75,000 tokens/day
- Priority analysis
- Export via MCP to Claude Code / Cursor
Enterprise
- Unlimited repositories
- Unlimited daily usage
- Private repos
- Multiple data sources
- Team seats & SSO