Not quite right? Say so in a comment.
Comment @builder keep the query string on the redirect too and the agent picks up the same session, on the same branch, and replies in the thread with what it changed.
Your AI software factory
Aictiq is the project board your coding agents can actually work from. Write a ticket, click Hand to agent, and Claude Code, Codex, OpenCode, Cursor or Copilot picks it up on your machine and sends back a pull request. No more copy-pasting tickets into a terminal.
No card. Open source (AGPL-3.0). Your code never leaves your machine.
Your system asks for less motion, so this page is standing still.
You, in 2026
You're not a developer anymore. You're a clipboard with opinions.
Aictiq takes steps 1 to 6. You keep the opinions.
1Write the ticket (or don't, really)
"Login redirect forgets the query string." That's a ticket. Paste a screenshot, drop a file, done.
Want it implementation-ready? Click Refine ticket. An agent reads the ticket, the screenshots and the relevant bits of your code, then rewrites it with acceptance criteria. If it can't guess something, it asks. You answer in the ticket, not in a chat window you'll lose.
#ACME-123 links with autocomplete2Hand it over
Hand to agent. Pick the playbook (your team's instructions, versioned in the wiki), pick the agent, optionally pick the machine. Or Start later: we prefill six hours from now, because we know what you're going to do tonight. Sleep.
Prefer automation? Add a rule: "when a ticket with label agent-ok enters Ready, run Implement as builder". The board becomes the queue.
3Your runner picks it up
The runner is one command on your laptop, a VPS or a CI box.
It only makes outbound HTTPS calls. No open ports, no VPN, no "please whitelist our IP range". It asks for work, we answer. Your repository, your harness subscription, your rules. Aictiq dispatches and records. It never holds your code.
$ npm install -g @aictiq/cli
$ aictiq runner register --url https://app.aictiq.com --token jrn_…
$ aictiq runner startKeep scrolling. The ticket does the travelling.
4The agent works (and you can watch)
The runner makes a fresh worktree, hands the harness a token that only lives as long as this run, and the agent gets to work. It reads the ticket over MCP, keeps one progress comment updated (not 40 of them), runs your tests and pushes a branch.
The live log streams into Aictiq as it happens. Close the tab. Your phone will tell you when it's done, via email, Slack, Discord or Telegram.
5Pull request, ticket moves itself
When the run ends, Aictiq links the pull request, writes one outcome comment, releases the ticket and moves it to the state your playbook says: In Review on success, something else on failure. Nobody drags cards around at 6 pm anymore.
Small project, no PR ceremony? Set the playbook to push directly to the default branch. We won't judge. Your branch protection might.
6You're still the boss
Comment @builder keep the query string on the redirect too and the agent picks up the same session, on the same branch, and replies in the thread with what it changed.
Continue resumes the agent's session where it stopped, with its edits still there. Retry starts clean.
Copy resume command gives you the exact claude --resume … line (or Codex, OpenCode, Cursor, Copilot) for that checkout.
Merge stays a human button.
That's not a missing feature, that's the point.
| Question | The usual setup | Aictiq |
|---|---|---|
| Where the agent runs | In a terminal tab you opened and forgot | On a runner you registered, as a service |
| Who picks the ticket | You, by copy-paste | The agent claims it. One claim per ticket, enforced by the database |
| Where its progress lives | Scrollback | One progress comment on the ticket, plus the live log |
| When two agents start at once | Two PRs, both wrong | One claim, one polite refusal |
| On a rate limit | You notice an hour later | Auto-continue after 1 and 5 minutes |
| Who did what | "The AI" | A named agent, owned by a named person, in every commit and comment |
| Whose code, whose machine | Theirs, often | Yours. Always |
| Which agent | The one the vendor sells | Claude Code, Codex, OpenCode, Cursor, Copilot. Switch per run |
It's also just a really good tracker
Boards, backlog, sprints, epics down to tasks, a wiki, analytics, keyboard shortcuts, a command palette. Fast at 100,000 items. Teams that only track work can ignore the AI bits forever. Nothing nags.
Proof
Every change since the public release is an Aictiq ticket, refined, run and reviewed in Aictiq. The commit log reads AICTIQ-54 Add GitHub Copilot CLI as a fifth runner harness. That ticket was handed to an agent. So was this one, probably.
Counted from GitHub at build time, 2026-10-07.
No. The runner works on your machine. Aictiq gets the status, the log stream, the PR link and the summary.
Claude Code, Codex, OpenCode, Cursor and GitHub Copilot via the runner. Any MCP client for the bring-your-own-agent loop.
For PRs, the runner uses your git and gh login, or the GitHub App. Direct-push mode works with any origin remote.
No, on purpose. It produces reviewable code. Your pipeline deploys.
They claim one ticket at a time, keep one progress comment, and Aictiq moves the ticket at the end, not the agent.
Yes, with the Stakeholder preset, and only what you want them to see.
Yes, AGPL-3.0. The name and logo are trademarked.
Absolutely. It's a full tracker first.
No card, no automatic charge. Or self-host everything for free, forever.
Opened PR #142. Tests, lint and type-check pass. Moved to In Review and released the ticket.