Free and open source. In development, no public release yet.
Your terminal.Your AI.Your workflow.
Code with your favorite AI, manage tasks, review pull requests, and run coding agents, all in one terminal.
Product preview. An illustration of the intended experience, not a recording of the application.
A ticket from your Board, taken to a reviewed pull request. The agent plans, you approve, it builds and tests on its own branch, a second agent checks the work, and nothing merges until you say so.
A better terminal
A modern terminal for everyday development. It works with no AI connected, and it stays fast when AI is slow.
More about the terminalTabs and splits
Arrange shells the way you work. Splits never restart a shell.
SSH built in
Open remote shells with host keys you have checked.
Search and restore
Find anything in scrollback. Pick up where you left off after a restart.
Your shell, your keys
Bash, Zsh, Fish, PowerShell or WSL, with themes and keybindings you set.
Bring your own AI
Use the models you already trust. No Vekrun subscription, no lock-in to one provider, and nothing is sent anywhere without you seeing it first.
Hosted providers bill you at their own rates. Local models are free to run.
More about AI codingYour API key
Connect Claude, OpenAI, Gemini or any OpenAI-compatible endpoint. You pay the provider directly, and keys stay in your OS keychain.
- Anthropic
- OpenAI
- Google Gemini
- OpenAI-compatible
A local model
Use Ollama or LM Studio and nothing leaves your machine.
- Ollama
- LM Studio
A coding-agent account you already pay for
PlannedPlanned. Only through each provider’s supported sign-in, and never required.
- Claude Code
- Codex
- Gemini CLI
Your tasks, ready to build
Connect Jira, Linear or Asana. See the tickets assigned to you. Click Start Work and you have a branch, a worktree and a terminal, with the ticket as context for you or an agent.
- Connect your tracker
- See your assigned tickets
- Click Start Work
- Start coding
- VEK-142Li
Add GitHub OAuth login
- API-88Ji
Rate-limit the export endpoint
- WEB-311Gh
Fix the dark-mode flash on load
- VEK-137Li
Migrate sessions to signed cookies
- OPS-21As
Rotate the staging database password
- VEK-129Li
Replace the retry queue
- API-80Ji
Return 429 with a Retry-After header
After Start Work, choose how to work on VEK-142
Vekrun creates the branch and worktree and opens a terminal in it. You write the code. The ticket is on your Board; the context is on your machine.
The ticket becomes the agent’s brief. It plans, edits and runs tests in the worktree, showing every diff before it writes and every command before it runs.
For larger work, a plan is drafted for your approval, agents implement in parallel on their own branches, a separate agent checks the result, and you approve the merge.
Product preview. The Board works with local tickets in the development build; connecting Linear, Jira, Asana, GitHub and GitLab is planned.
Your tasks, your terminal, and your AI, all connected.
AI coding teams
For work that is bigger than one conversation: agents plan it, implement and test it on separate branches, a different agent reviews the result, and you keep the final say.
Start from a ticket on your Board or a sentence of your own. Say what done looks like.
What you decide
- Edit the goal and acceptance criteria
- Pick which models to use and how much to spend
An agent reads the project and drafts a plan: steps, files, dependencies, risks and the tests that will prove it.
What you decide
- Edit any step
- Approve the plan, or send it back
Optional. Several models evaluate the plan independently and challenge each other’s recommendations. Disagreements are shown, not hidden.
What you decide
- Choose the members, models and rounds
- Skip it for routine work
Agents implement the approved steps in parallel, each on its own branch and worktree, so nothing collides.
What you decide
- Approve diffs and commands as they come
- Pause, redirect or take over any task
The tests you allow run in each worktree. Then a different agent checks the result against the goal, with the commands and output as evidence.
What you decide
- Decide which commands run without asking
- Accept the check or ask for more
You see a merge preview with any conflicts. Nothing is merged, pushed or updated on your tracker until you approve.
What you decide
- Approve the merge, or not
Every stage is optional. Small tasks go straight to a single agent, and a plan review by several models is there when a decision matters.
Code reviews in one place
Pull requests, AI findings and your feedback, side by side with the code.
Planned The pull-request inbox is planned. AI review of local branches works today.
More about code reviews- See the pull requests waiting for your review.
- Open a teammate’s changes and read the diff where you code.
- Ask AI to point out problems before you do.
- Compare findings from several models when it matters.
- Leave your feedback without leaving Vekrun.
Free and open source
No account, no subscription, no hosted service in the middle.
Runs on your machine
Sessions, memory and credentials stay local. You can see what is sent to a provider before it goes.
You set the permissions
Agents write files and run commands only when you allow it. Why we build it this way.
Your development environment should belong to you.
One terminal. Any supported model. Your tasks and reviews where you write the code.
- TerminalA modern terminal for everyday work, with tabs, splits, SSH and search.
- AI codingCode with Claude, OpenAI, Gemini or a local model, on your terms.
- BoardYour assigned tickets, one click from a branch, a terminal and an agent.
- AI TeamsPlan, build, test and verify larger work with coordinated agents.
- Code reviewsPull requests, AI findings and your feedback in one place.
Vekrun is in development. There are no public releases yet. Status as of 2026-10-19.