Short answer: open Cody's chat panel in your IDE, press ⌃+⌥+R (Mac) or Ctrl+Alt+R (Windows/Linux), speak the question with full codebase context, and AICHE drops a clean prompt into Cody. Cody searches the repo, you get a grounded answer.
Cody Knows Your Whole Codebase. Your Prompts Should Too.
Sourcegraph Cody indexes your entire repository. It can follow imports across services, find similar patterns in distant files, and ground its answers in your actual code, not generic snippets from training data. That advantage only kicks in when your prompt names the things Cody can search for.
"How do I add caching" gives Cody nothing.
"Add Redis caching to UserService following the cache-aside pattern we use in OrderService, with the invalidation strategy from the event-bus topic" gives Cody enough to find the existing patterns and produce code that fits.
The detailed prompt takes 5-10 minutes to type. It takes 30-45 seconds to speak. That's the gap AICHE closes.
How It Works
- Open your IDE (VS Code, JetBrains, or Neovim) with the Cody extension installed.
- Open Cody's chat panel in the sidebar.
- Press ⌃+⌥+R (Mac) or Ctrl+Alt+R (Windows/Linux) to start.
- Speak the full question with service names, file references, and constraints. No length cap.
- Press the hotkey again. AICHE transcribes, cleans up, inserts the prompt.
- Hit Enter. Cody searches your repo and answers.
Where Voice Pays Off in Cody
Architectural Walk-Throughs
"Walk me through what happens when a user clicks checkout, starting from the React component, through the API gateway, the payment service, and the notification system. Show me the actual code path." Typed: 4 minutes. Spoken: 20 seconds. Cody traces it through your repo because the prompt names every layer it should look at.
Pattern-Matched Code Generation
"Create a bulk user import endpoint following the same pattern as the bulk order import in order-service: same validation, same accumulate-errors-and-return strategy, same WebSocket progress notifications." Cody finds the existing pattern, copies the shape, generates code that matches your conventions instead of inventing new ones.
Onboarding and Code Review
Speaking is faster than reading. Pace through a feature or PR, dictating questions as they come up: "Why does the retry logic in the queue consumer use exponential backoff but the API client uses fixed intervals?" Cody pulls both files and explains the divergence.
Refactor Planning
"List every place in the repo that imports from the legacy userv1 module, group them by service, and tell me which ones already have v2 equivalents." That kind of cross-cutting question is what Cody is built for, and the spoken prompt gets you there in seconds.
What You Get
- Unlimited voice notes with AI cleanup - filler words removed, punctuation and paragraph breaks added.
- Software Development profile (Pro) - recognition tuned for code, APIs, library names, CLI flags.
- Custom vocabulary - feed AICHE your service names, internal libs, project codenames.
- System-wide dictation - one hotkey, every IDE, every text field.
- Multilingual voice input - speak in your native language, transcribe in that language or auto-translate to English.
- Zero-retention audio - audio purged immediately after processing, within 1 second.
Plans start at $3.99/mo (annual) with a 7-day free trial, no credit card. See pricing.
Common Questions
Q: Does it work in JetBrains IDEs as well as VS Code?
A: Yes. AICHE inserts wherever your cursor is, including the Cody chat panel in IntelliJ, PyCharm, GoLand, RubyMine, WebStorm, etc.
Q: I work on a monorepo with internal package names that aren't in the dictionary. Will Cody understand them?
A: Add them to AICHE's Custom Vocabulary so they get spelled correctly in the prompt. Cody searches the repo for those exact strings, so spelling matters.
Q: Can I dictate the codebase-context hints (@file, @symbol)?
A: Yes. Say "at file" or "at symbol" and AICHE transcribes the literal text. You can also turn on a custom rule in AICHE settings to substitute "at file" for @file automatically.
Q: My team uses Cody Enterprise. Anything different?
A: No. Voice input is independent of which Cody tier you run. The same hotkey works against the chat panel regardless.
Result: architectural questions that took 10 minutes to type now take a minute to speak, and Cody answers better because the prompt actually references your service names and patterns instead of vague generalities.
Try it now: open Cody, press your hotkey, and ask one question that spans more than three files. Reference services and patterns by their real names. See how Cody uses that context.