Paca is the first project tracker I have used that treats an AI agent as a teammate rather than a chatbot bolted onto Jira. For a solo founder running multiple parallel agents, the open-source Scrumban board and MCP server make sense. For a small team still doing mostly human work, plain Linear is faster.
I have tried roughly a dozen project trackers since I started running coding agents inside my own company, and almost all of them have the same shape: a normal Jira-style board with a chatbot welded onto the side. You can ask the chatbot to summarise a sprint, or to draft a ticket, but the agent never actually moves a card. It is, as the Paca team puts it on their landing page, a chatbot gap. The work still happens in the human columns; the AI is a peripheral. So when Paca launched on Hacker News this month as an open-source Jira alternative built around human-AI collaboration in Scrum, I spun up a local instance the same evening to see whether it is genuinely different or just the same chatbot wearing new branding.
The shift is real, and it is small but consequential. In Paca, an agent — Claude through the MCP server, in my test — gets assigned to a sprint exactly the way a person does. It shows up on the Scrumban board with its own avatar. It picks tasks from the backlog, moves them across the board, writes BDD scenarios, and contributes to system design documents. From a coordination point of view, that is the thing I have wanted for two years: a single source of truth where I can see what my agents are doing and intervene at the column boundary rather than scrolling through chat transcripts trying to reconstruct state. Self-hosting matters too; if you run anything client-sensitive or you are in the EU, having the project graph on your own machine is not optional any more.
But the honest part. There is a Dorie Clark line in The Long Game that I keep coming back to when I evaluate new tools: we overestimate what we can accomplish in a day and underestimate what we can accomplish in a decade. New tools tend to win the day and lose the decade because the unglamorous infrastructure they replace — Linear, GitHub Issues, plain markdown — has had ten years of compounding polish. Paca is at the start of its compounding curve. The Scrumban board is rough. The BDD authoring is fiddly. If your team is four humans and one occasional agent, the cost of switching is higher than the benefit. The category Paca belongs to (AI-native PM, alongside tools like Plane and Bridge) is also crowded, and the winner in two years is not obvious yet.
Where it earns its place, for me, is the specific shape of the one-person company. I run between three and six agents in parallel on a normal week — one writing copy, one fixing infra issues, one prototyping a feature, one mining content. The cognitive cost of tracking all of that in my head is high enough that it eats the leverage the agents are supposed to give me. Paca, even in its rough state, gives me one board where I can see all six. The MCP server means I do not have to teach each agent a separate API; they read and write the workspace the same way I do. That alignment between agent and human surface is the actual product, and it is what makes the comparison to Jira (or Linear, which I prefer for humans) misleading. Paca is not trying to be a faster Jira. It is trying to be the operating layer for a company where the workforce is mostly software.
Ethan Mollick describes the right mental model in Co-Intelligence: treat the AI as a person on the team, not a tool you query. Most software is still built around the second metaphor — a smart autocomplete you summon. Paca is built around the first, and the difference shows up in tiny moments. When an agent updates a card, the activity feed reads like a teammate's, not like a log line. When I assign work, I do not feel like I am queuing a job; I feel like I am delegating. That feeling matters because it is how I behave next — I check in less often, I write better tickets, I trust the agent further. The tool shapes the relationship.
The verdict I would give a founder asking me whether to migrate: do not switch your existing team to Paca yet. If you are mostly humans, the friction is not worth the future-proofing. But if you are a solo founder or a two-person team running multiple agents in parallel and your current workflow is already a mess of Slack threads and half-tracked AI work, install Paca this weekend and run one real project on it. The thing you are evaluating is not feature parity with Jira. You are evaluating whether you can finally see what your agents are doing and lead them as if they were people. If the answer after two weeks is yes, the rest of the product will improve under you. If the answer is no, you have lost a weekend, which against a decade-long bet on agent-led work is a rounding error.
Related: How to Find Your Passion · Best Self-Improvement Books · How to Make Better Decisions · What University Will Not Teach You