* docs: add issue template requirements and fix anchor links in CONTRIBUTING.md * fix: align feature request docs with actual template fields --------- Co-authored-by: Vishal Kumar Singh <singhvishalkr@gmail.com>
6.7 KiB
Contributing to Kilo CLI
See the Documentation for details on contributing.
TL;DR
There are lots of ways to contribute to the project:
- Code Contributions: Implement new features or fix bugs
- Documentation: Improve existing docs or create new guides
- Bug Reports: Report issues you encounter
- Feature Requests: Suggest new features or improvements
- Community Support: Help other users in the community
The Kilo Community is on Discord.
Developing Kilo CLI
-
Requirements: Bun 1.3.13+
-
Install dependencies and start the dev server from the repo root:
bun install bun dev
Developing the VS Code Extension
Build and launch the extension in an isolated VS Code instance:
bun run extension # Build + launch in dev mode
This auto-detects VS Code on macOS, Linux, and Windows. Override with --app-path PATH or VSCODE_EXEC_PATH. Use --insiders to prefer Insiders, --workspace PATH to open a specific folder, or --clean to reset cached state.
Running against a different directory
By default, bun dev runs Kilo CLI in the packages/opencode directory. To run it against a different directory or repository:
bun dev <directory>
To run Kilo CLI in the root of the repo itself:
bun dev .
Running Kilo CLI from any folder
bin/kilodev is a self-locating launcher that runs this checkout from wherever you invoke it. Running it with no arguments launches the TUI pointed at the caller's directory; any arguments are forwarded to the CLI unchanged.
One-shot install (recommended). From the repo root:
./bin/kilodev dev-setup
This detects your shell, shows exactly what it will add, asks for confirmation, writes an idempotent block to your rc file, and saves a timestamped backup of the original. Re-running is safe — it only rewrites when the snippet has changed.
Useful flags:
--yes— skip the confirmation prompt (good for CI/containers).--print— just print the snippet, don't touch any file (pipe-friendly).--dry-run— show what would change without writing.--shell <zsh|bash|fish|powershell>— override shell detection.--rc <path>— override the rc file.
Manual alternatives (equivalent, no CLI invocation needed):
- Unix: add
alias kilodev='/path/to/kilocode/bin/kilodev'to~/.zshrc/~/.bashrc, orfish_add_path /path/to/kilocode/bin. - Windows: add
C:\path\to\kilocode\binto PATH (System Environment Variables), or addfunction kilodev { & "C:\path\to\kilocode\bin\kilodev.cmd" @args }to$PROFILE.
Then from anywhere:
cd ~/some/project
kilodev # opens TUI with project = ~/some/project
kilodev dev-setup --print # prints the alias line (scripting)
kilodev run --dir "$PWD" "…" # subcommands pass through; use --dir for run/serve
Building a "local" binary
To compile a standalone executable:
./packages/opencode/script/build.ts --single
Then run it with:
./packages/opencode/dist/@kilocode/cli-<platform>/bin/kilo
Replace <platform> with your platform (e.g., darwin-arm64, linux-x64).
Understanding bun dev vs kilo
During development, bun dev is the local equivalent of the built kilo command. Both run the same CLI interface:
# Development (from project root)
bun dev --help # Show all available commands
bun dev serve # Start headless API server
# Production
kilo --help # Show all available commands
kilo serve # Start headless API server
Testing with a local backend
To point the CLI at a local backend (e.g., a locally running Kilo API server on port 3000), set the KILO_API_URL environment variable:
KILO_API_URL=http://localhost:3000 bun dev
This redirects all gateway traffic (auth, model listing, provider routing, profile, etc.) to your local server. The default is https://api.kilo.ai.
There are also optional overrides for other services:
| Variable | Default | Purpose |
|---|---|---|
KILO_API_URL |
https://api.kilo.ai |
Kilo API (gateway, auth, models, profile) |
KILO_SESSION_INGEST_URL |
https://ingest.kilosessions.ai |
Session export / cloud sync |
KILO_MODELS_URL |
https://models.dev |
Model metadata |
VS Code: The repo includes a "VSCode - Run Extension (Local Backend)" launch config in
.vscode/launch.jsonthat setsKILO_API_URL=http://localhost:3000automatically.
Issue Template Requirements
If you open an issue through the GitHub web UI, GitHub will guide you through the correct template automatically.
If you open an issue through gh issue create, the API, or another tool that bypasses the web UI, include the equivalent required fields yourself so the issue still matches the template. Issues that skip required fields may be auto-closed by the compliance bot.
Current required fields by issue type:
- Bug report: include a
Description. When you can, also add Plugins, Kilo version, Steps to reproduce, Screenshot and/or share link, Operating System, and Terminal so the report matches the full bug template. - Feature request: use a title prefixed with
[FEATURE]:, complete the required checkbox confirming you have searched for duplicates, and fill inDescribe the enhancement you want to request. - Question: include the
Questionfield.
Pull Request Expectations
- UI Changes: Include screenshots or videos (before/after).
- Logic Changes: Explain how you verified it works.
Issue First Policy
All pull requests must reference an existing issue.
This helps reviewers understand the problem statement, discussion, and intended scope before reviewing the code change.
PR Titles
Use conventional commit style PR titles such as:
feat: add MCP settings tabfix: correct Windows path handlingdocs: clarify issue template requirementschore: bump TypeScript to 5.8refactor: extract diff renderer into a hooktest: cover ServerManager orphan cleanup
Issue and PR Lifecycle
To keep our backlog manageable, we automatically close inactive issues and PRs after a period of inactivity. This isn't a judgment on quality — older items tend to lose context over time and we'd rather start fresh if they're still relevant. Feel free to reopen or create a new issue/PR if you're still working on something!
Style Preferences
- Functions: Keep logic within a single function unless breaking it out adds clear reuse.
- Destructuring: Avoid unnecessary destructuring.
- Control flow: Avoid
elsestatements; prefer early returns. - Types: Avoid
any. - Variables: Prefer
const. - Naming: Concise single-word identifiers when descriptive.
- Runtime APIs: Use Bun helpers (e.g.,
Bun.file()).