A lot of project-management tools were designed around humans creating and completing tasks.
But when you’re building with AI agents, that model starts to feel different.
An agent might handle data preparation, another could run experiments, while a researcher reviews outputs and decides what should happen next.
The challenge becomes coordinating humans and agents, not just tracking tasks.
I’ve been exploring Sharkly.ai as a free AI-native approach to this problem. Instead of treating agents as separate assistants, the idea is to make them part of the project workflow alongside human teammates.
For teams building with Hugging Face models and agent workflows, I’m curious:
Would you want your task-management tool to understand AI agents as actual contributors to the project?
Or should project management and agent execution remain separate?
Short answer: agents need to be first-class assignees with a visible action log and a hard “waiting for human review” state, but that doesn’t require a separate task board. Plain issue trackers (GitHub Issues/Projects, Linear) plus an agent framework’s own trace logging already cover most of it. What actually matters is ownership per task, an audit trail of what the agent did, and an explicit human approval gate. Everything else is UI.
Also worth flagging: this is the same product (Sharkly.ai) that’s come up in several of your recent threads here, plus similar posts on GitHub Discussions and dev.to. It’d be more useful to just say upfront that you’re promoting it, or to share concrete details (pricing, limits, what it does that Linear plus an agent framework doesn’t). Otherwise people will read these as ads rather than open questions.