Project work for people and AI agents

Your team and its AI agents, working from the same records.

TaskScoop gathers what comes in, turns it into tasks, and keeps what your team learns. Soon, when you correct an agent, the correction sticks.

Five regions + one loop

Capture

Live

Execution

Live

Knowledge

Coming

Procedures

Coming

Rules

Live

Correction loop

Coming

Concept

Everything has one home, so nothing gets lost and agents never guess.

Project information is scattered across chats, files and people's heads. TaskScoop gives each kind of information one place: what came in, what we are doing, what we know, how we do it, and who may act. AI agents are members of your organization and work inside the same places, under the same rules.

Five regions

Five places, each with one job and one rule, plus a loop that learns.

Capture

Files, chats and connector items come in with their source. People and agents never change them, so work can always point back to where it came from.

Execution

Goals, tasks and decisions, with clear owners. Edit together in realtime, even offline. AI agents join as members of your organization.

Coming

Knowledge

Short pages of what your team knows, each citing its sources. A newer page replaces the old one, and the old one stays linked.

Coming

Procedures

How your team does recurring work: steps and checks, promoted only after they have worked on real tasks.

Rules

Roles and permissions for people and agents, with agent instructions. Agents only act within the rules you set. Approval gates are coming.

Coming

Correction loop

When you fix an agent's work, the fix updates knowledge or a procedure, so the next run starts from it.

Correction loop ยท Coming

Where we are going: every correction makes the next run better.

01

Agent drafts

An agent will propose work on a task, using your knowledge and procedures.

02

You correct

You accept, edit or discard. Nothing is applied without a person.

03

TaskScoop learns

Your fix is saved to a knowledge page or procedure. TaskScoop learns from records, not from model training.