Tuskway vs an issue tracker alone
Your tracker knows the status. Your agents need the plan.
An issue tracker answers "who is doing what, and is it done?". It does that well, and Tuskway does not try to replace it. What a tracker does not hold is the reasoning: why this approach, what was ruled out, which files a ticket touches, what "done" means in a way an agent can check. Teams that hand a bare issue to a coding agent get code that matches the title and misses the intent.
Side by side
Tuskway vs an issue tracker alone
| an issue tracker alone | Tuskway | |
|---|---|---|
| What it holds | Titles, descriptions, status, assignee. | The context, a sectioned tech design, and a brief per ticket with acceptance checks. |
| Written for | People who were in the discussion. | People and agents who were not. |
| Review | Comments on the issue, after it exists. | Line comments on the design, before tickets are cut. |
| What an agent reads | Whatever fits in the issue description. | The brief, plus the one design section the ticket comes from. |
| Dependencies | Links between issues, set by hand. | Dependencies and touched areas on each ticket, used to show what is ready and what would collide. |
| When the issue is edited | The description changes. | The change is flagged next to the brief and an editor decides which version to keep. |
| History | Activity log on the issue. | Every version of every note, with who changed it, person or agent. |
What a tracker is good at
Trackers are built for flow: boards, cycles, assignment, reporting. Everyone on the team, technical or not, can see where a feature stands. None of that needs to move. If your tickets are small, the work is familiar, and one person does it by hand, an issue with two lines is enough.
Where it stops for agents
An agent starts every session knowing nothing about your feature. It will read the issue it is given and fill the gaps with guesses. The gaps are the expensive part: the constraint someone raised in a meeting, the interface two tickets have to agree on, the approach you already tried. Pasting all of that into each issue description by hand does not scale, and long descriptions drift as soon as the plan changes.
How the two work together
In Tuskway the feature is planned first: your agent drafts the context and the tech design, the team comments on the lines that need work, and the design is split into tickets. One click publishes those tickets to GitHub Issues, GitLab, Linear or Jira. From then on status, assignee and labels sync both ways. If someone rewrites an issue in the tracker, Tuskway does not overwrite it and does not silently accept it: it shows the difference and asks.
A worked example
Take "retry failed card payments". In a tracker alone, the issue says exactly that. In Tuskway, the ticket's brief says: retry after 1 minute, 10 minutes and 1 hour; one idempotency key per attempt; stop after three; mark the order payment_failed; touch the payments module only. It links to the design section where a reviewer asked for retry keys to expire after 24 hours, and the agent's answer. The published issue carries the brief. The agent that picks it up builds what the team agreed, not what the title suggests.
Choose an issue tracker alone when
Tickets are small and routine, one person does the work, and nobody hands them to an AI agent.
Choose Tuskway when
Features span several tickets, several people or agents build them, and the plan needs review before the code.
Do we have to stop using our tracker?
No. Keep it. Tuskway publishes tickets to it and keeps them in step. See the pages for Linear, Jira, GitHub Issues and GitLab.
Can we use Tuskway without a tracker?
Yes. Connecting one is optional. Tickets, their status and the plan of what is ready all work inside Tuskway.
Who can publish tickets?
Anyone who can edit the epic, and their agents. A workspace owner connects the tracker once.
Try the team version of spec-driven development.
Connect your agents, run /start-epic on your next feature, and review the plan before anyone writes the code.