José Almeida 123b2f6eed feat: start task from GitHub URL (#365)
* feat(backend): accept github_url in task creation and clone on launch

Add github_url field to task creation DTO and propagate through HTTP/WS
handlers and service layer. When a task is created with a GitHub URL
instead of a local repository, the service resolves (or creates) the
repository record via FindOrCreateRepository and triggers a clone/fetch
via repoclone.Cloner during executor pre-flight.

Wire RepoCloner into the orchestrator executor so remote repositories
are fetched before the agent process starts. On resume, RepoUpdater
performs a fetch to ensure the worktree is up-to-date.

* feat(backend): add GitHub branches endpoint for remote repos

Add ListRepoBranches to the GitHub client interface (GHClient,
PATClient, NoopClient, MockClient) and expose it via
GET /api/v1/github/repos/:owner/:repo/branches.

The mock controller gains POST /api/v1/github/mock/branches for
seeding branch data in E2E tests.

* feat(web): add GitHub URL input to task create dialog

Add a toggle in the task create dialog that switches between the
existing repository selector and a free-text GitHub URL input. When a
valid GitHub URL is entered (owner/repo parsed), branches are fetched
from the new backend endpoint and auto-selected.

The dialog state, effects, validation, and submit logic are extended to
support the github_url flow alongside the existing repository flow.

* test(e2e): add tests for task creation from GitHub URL

Add E2E tests verifying:
- Full flow: toggle to GitHub URL mode, enter URL, create task, verify
  agent runs and completes in session view
- Toggle behavior: switching between GitHub URL and repository selector
  modes updates UI correctly

* fix(web): dynamic URL input sizing and extract RepoSourceInput

Replace fixed-width GitHub URL input with size-based auto-sizing
(matching InlineTaskName pattern). Extract RepoSourceInput component
from DialogHeaderContent to stay within cyclomatic complexity limits.

Add testId prop to Combobox and wire it to ExecutorProfileSelector
for E2E test targeting.

* test(e2e): add worktree executor test and fix GitHub URL repo seeding

Update GitHub URL E2E tests to pre-seed a repository with GitHub
provider info pointing to the local git repo, avoiding actual clone
attempts during task creation.

Add test for GitHub URL task creation with worktree executor type,
verifying the full flow including executor profile selection.

Extend ApiClient with listExecutors and provider fields on
createRepository for test seeding.

* fix(backend): use protocol-aware clone URL for provider-backed repos

The ensureRepoCloned function was hardcoding HTTPS clone URLs via
repositoryCloneURL(), but the user's git setup may require SSH.

Add BuildCloneURL to the RepoCloner interface so the executor uses
the Cloner's configured protocol (SSH/HTTPS) instead of always
generating HTTPS URLs. Falls back to HTTPS if BuildCloneURL fails.

This fixes workspace loading issues (terminal, file tree) for cloned
repos when the user has SSH-only authentication configured.

* test(e2e): enhance GitHub URL test with terminal and changes panel verification

* feat(web): add inline validation for GitHub URL input

- Show inline error for invalid URL format (not matching github.com/owner/repo)
- Show inline error when repository is not found or not accessible
- Mock client returns error for unknown repos (matches real GitHub API)
- E2E tests for invalid URL, nonexistent repo, and error-clearing on valid input

* fix(web): fix GitHub URL branch auto-select and E2E test stability

- Pick branch by name (main/master) for GitHub URL mode instead of
  using selectPreferredBranch which returns origin-prefixed names
- Add test timeout for worktree executor test
- Use shorter filename in terminal test to avoid xterm line-wrap issues

* fix(web): fix GitHub URL dialog polish and wrong-repo bug

UI fixes:
- Align toggle button text with dropdown text content (pl-3)
- Match GitHub URL input height to dropdown trigger (min-h-9)
- Position error text absolutely to prevent layout shift
- Remove redundant toast on repo-not-found (inline error suffices)

Bug fix:
- Prevent repo auto-select from overriding GitHub URL mode by
  guarding useRepositoryAutoSelectEffect with useGitHubUrl
- Reorder getRepositoriesPayload to check useGitHubUrl first,
  preventing stale repositoryId from winning over the GitHub URL

Add E2E test verifying correct repo usage across consecutive tasks.

* fix(web): fix dialog height mismatch and error text overlap

- Use exact h-9 on GitHub URL input to match dropdown trigger height
- Move error text from absolute-positioned overlay to inline next to
  the toggle button, preventing overlap while keeping layout stable

* fix(web): make GitHub URL input auto-size to content

Reduce minimum size from 30 to 20 and shorten placeholder so the
"/ task-name" sits closer to the URL text, matching the dropdown's
compact w-auto behavior.

* fix(web): polish GitHub URL input sizing, floating error, and placeholder color

- Match input height (h-7) and padding (px-2) to Combobox trigger button
- Size input width to exactly fit placeholder text
- Show validation error as floating popover below input instead of inline
- Keep placeholder text grey even when input has error state
- Extract useGitHubUrlState hook to fix max-lines-per-function lint

* fix(backend): return proper HTTP status codes from GitHub branches endpoint

Add GitHubAPIError type to distinguish upstream GitHub API status codes.
The branches endpoint now returns 404 (not 500) for non-existent repos,
which the logging middleware correctly logs as DEBUG instead of ERROR.

* feat(web): show error toast when a task session fails to start

Add session failure notification plumbing: WS handler detects FAILED
state with error_message, pushes to Zustand store, and a global bridge
component shows an error toast via the ToastProvider. Also adds
data-testid attributes to toast elements and types error_message on
the session state changed payload.

* style: fix prettier formatting in task-create-dialog and related files

* fix(web): extract ToastList component to fix max-lines-per-function lint

* fix: address code review feedback for GitHub URL task creation

- Preserve 403 status instead of collapsing to 401 in branches endpoint
- Use clone-generic error instead of remote-docker-specific in ensureRepoCloned
- Trim whitespace in GitHub URL validation (backend + frontend)
- Add GitHubURL field to wsUpdateTask path
- Fix DefaultBranch mapping to prefer repoInput.DefaultBranch over BaseBranch
- Clear githubBranchesLoading on early-return paths to prevent stuck state
- Gate GitHub branch loading by mode to avoid false loading states
- Encode owner/repo in API path for special characters
- Add aria-label, aria-invalid, aria-errormessage for accessibility

* fix(e2e): improve diff-update test stability with fresh locator and longer timeout

* fix(backend): add --paginate flag to gh api branches call for repos with >100 branches

* fix: address second round of code review feedback

- Add pagination support to PATClient.ListRepoBranches for repos with >100 branches
- Validate owner/repo in mock controller addBranches endpoint
- Use aria-describedby instead of aria-errormessage for broader AT support
- Add aria-live region to toast container for screen reader announcements
- Assert absence of FIRST_MODIFICATION in diff-update E2E test

* fix(backend): downgrade HTTP request logger from error to warn for 4xx/5xx responses

* fix: address third round of code review feedback

- Validate repositoryID/localPath in UpdateRepositoryLocalPath
- Copy slices in mock client to prevent external mutation
- Extract GitHub header constants and shared setter in PATClient
- Redact clone URL from logs to prevent credential leaks
- Guard against persisting empty localPath after clone
- Show fallback toast message when FAILED state has no error_message
2026-03-05 00:47:58 +00:00
2026-02-26 23:08:15 +00:00
2026-03-02 08:54:58 +00:00
2026-01-16 01:59:19 +00:00
2026-03-05 00:43:02 +00:00
2026-02-17 09:19:42 +00:00
2026-03-02 08:54:58 +00:00

Kandev

Manage and run tasks in parallel. Orchestrate agents. Review changes. Ship value.

Workflows | Roadmap | Contributing | Architecture

What

Organize work across kanban and pipeline views with opinionated workflows and execute multiple tasks in parallel. Assign agents from any provider, and review their output in an integrated workspace - file editor, file tree, terminal, browser preview, and git changes in one place. Terminal agent TUIs are great for running agents, but reviewing and iterating on changes there doesn't scale.

Run it locally or self-host it on your own infrastructure and access it from anywhere via Tailscale or any VPN.

Open source, multi-provider, no telemetry, not tied to any cloud.

Vision

Humans stay in control. Define tasks, build agentic workflows with gates, review every change, decide what ships.

  • Review-first - Humans support production systems. We need to understand (yet) and trust the code that gets deployed.
  • Your workflow - Every team is different, and not every developer uses AI the same way. Define workflows once, share them across the team, and give everyone a consistent process for working with agents - regardless of experience level.
  • Remote agents - Running multiple agents on a large codebase can quickly saturate a local machine. The goal is a single control plane: offload execution to servers, orchestrate from anywhere, including your phone.
Where this is heading

While agents can already plan, implement, test, and deploy autonomously with the right workflows, we don't yet trust their output enough for production-grade software without human review. Crossing that trust threshold will fundamentally transform our industry.

Features

  • Multi-agent support - Claude Code, Codex, GitHub Copilot, Gemini CLI, Amp, Auggie, OpenCode
  • Parallel task execution – start and manage multiple tasks from different sources simultaneously, boosting efficiency and productivity with AI agents.
  • Integrated workspace - Built-in terminal, code editor with LSP, git changes panel, and chat in one IDE-like view
  • Kanban task management - Drag-and-drop boards, columns, and workflow automation
  • Agentic workflows - Multi-step pipelines that chain agents through automated task routing. See docs/workflows.md
  • CLI passthrough - Drop into raw agent CLI mode for direct terminal interaction with any supported agent, leverage the full power of their TUIs
  • Workspace isolation - Git worktrees prevent concurrent agents from conflicting
  • Flexible runtimes - Run agents as local processes or in isolated Docker containers
  • Session management - Resume and review agent conversations
  • Stats - Track your productivity with stats on the completed tasks, agent turns, etc

Supported Agents

Agent Default Model Other Models Protocol
Claude Code Sonnet 4.5 Opus 4.6, Opus 4.5, Haiku 4.5 stream-json
Codex GPT-5.2 Codex GPT-5.1 Codex Max, GPT-5.1 Codex Mini, GPT-5.2 Codex
GitHub Copilot GPT-4.1 GPT-5.2, Claude Sonnet 4.5, Gemini 3 Pro, +10 more Copilot SDK
Gemini CLI Gemini 3 Flash Gemini 3 Pro ACP
Amp Smart Mode Deep Mode stream-json
Auggie Sonnet 4.5 Opus 4.5, GPT-5.1, GPT-5, Haiku 4.5 (dynamic) ACP
OpenCode GPT-5 Nano Claude Sonnet 4, Opus 4, etc (dynamic) REST/SSE

CLI Passthrough mode - we should be able to support any agent that has a CLI interface, even if they don't have an official API or SDK. The agent runs in a terminal session and we pass input/output through a websocket connection. If your agent isn't supported yet, open an issue or submit a PR with the integration. See Adding a New Agent CLI for a step-by-step guide.

Quick Start

Beta - Under active development. Expect rough edges and breaking changes. Contributions welcome! See CONTRIBUTING.md for details.

npx kandev

This downloads pre-built backend + frontend bundles and starts them locally. The worktrees and sqlite db will be created in ~/.kandev by default. Should work on macOS, Linux, and Windows (WSL or native).

From Source

# Clone the repository
git clone git@github.com:kdlbs/kandev.git
cd kandev

# Start in production mode
make start

Prerequisites: Go 1.26+, Node.js 18+, pnpm, Docker (optional - needed for container runtimes)

High level architecture

graph LR
    UI[Web UI] --> Backend


        Backend["Backend (Orchestrator)"]


    Backend --> W1 & W2

    subgraph W1[Local Process]
        Agent1[Agent CLI] --- WT1[Worktree]
    end

    subgraph W2[Docker Container]
        Agent2[Agent CLI] --- WT2[Worktree]
    end

We also want to add support for these remote runtimes:

  • Remote SSH - run agents on remote servers over SSH, using docker or local processes with workspace isolation
  • K8s operator - run agents in a Kubernetes cluster, with auto-scaling and resource management.
Development

Project Structure

apps/
├── backend/    # Go backend (orchestrator, lifecycle, agentctl, WS gateway)
├── web/        # Next.js frontend (SSR, Zustand, real-time subscriptions)
├── cli/        # CLI tool (npx kandev launcher)
└── packages/   # Shared UI components & types

Prerequisites

  • Go 1.21+
  • Node.js 18+
  • pnpm
  • Docker (optional)

Running Dev Servers

# Start everything (backend + frontend with auto ports)
make dev

# Or run separately
make dev-backend    # Backend on :8080
make dev-web        # Frontend on :3000

Testing & Linting

make test           # Run all tests (backend + web)
make lint           # Run all linters
make typecheck      # TypeScript type checking
make fmt            # Format all code

Pre-commit Hooks

# Install pre-commit (https://pre-commit.com/#install)
pipx install pre-commit

# Install git hooks
pre-commit install

Comparison to Other Tools

There are a few similar tools in this space, and new ones appearing everyday. Here's what sets this one apart:

  • Server-first architecture - Not a desktop app. Runs as a server you can access from any device, including your phone. Start a task away from your computer and check in on it later.
  • Remote runtimes - Run agents on remote servers and Docker hosts, not just your local machine.
  • Multi-provider - Use Claude Code, Codex, Copilot, Gemini, Amp, Auggie, and OpenCode side by side. Not locked to one vendor.
  • CLI passthrough and chat - Interact with agents through structured chat messages or drop into raw CLI mode for full agent TUI capabilities.
  • Open source and self-hostable - No vendor lock-in, no telemetry, runs on your infrastructure.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md before opening a PR.

See the issue tracker for open tasks.

Acknowledgments

Built with these excellent open-source projects:

Monaco Editor · Tiptap · xterm.js · dockview · CodeMirror · dnd-kit · Mermaid · Recharts · TanStack Table · Zustand · Shadcn/UI · Radix UI · Tailwind CSS

License

AGPL-3.0

S
Description
Upstream mirror of https://github.com/kdlbs/kandev (main). Internal development and testing.
Readme 371 MiB
Languages
Go 58.3%
TypeScript 37.6%
HTML 1.9%
JavaScript 0.8%
Python 0.7%
Other 0.5%