Built for more than one.
Multiple people. Multiple agents. One shared memory, task ledger, and coordination layer — so the team moves as one system instead of a collection of parallel chat windows.
wikiTaTa is built for the way AI-assisted work actually happens: across multiple people, multiple sessions, and multiple agents running simultaneously. The same structured memory, tasks, and state that makes a solo project coherent becomes the shared coordination layer when a team is involved. There is no separate "team edition" — multi-user is the default.
How it works
Shared knowledge from the first card
Every card, task, note, and decision is scoped to the project and visible to every participant. A new teammate — human or AI — gets the full picture by reading the cards, not by scheduling a briefing or sifting through chat history. The knowledge base is the onboarding document.
Agent-to-agent directives
One Claude can assign a structured task to another Claude and receive the result back. This is the Directive system: a turn-based, typed request that travels between agents with a reply path built in. One agent sends a job; the other picks it up, executes it, and posts the result — every step logged and confirmed. It is one of two live coordination channels; the other is the Now Board intercom described below.
Live presence and collision prevention
wikiTaTa tracks who is active right now — which humans are online and which agents are in session — so two participants never unknowingly pick up the same task. Work assignments and in-progress state are visible across the whole team before anyone starts, preventing duplicated effort before it happens.
Cross-session messaging
Direct messages travel between sessions, with priority levels. An urgent message can interrupt an active session; a normal one waits in the inbox. Either way, the reply thread stays attached to the original — communication is structured, not lost in a side chat.
Session handoffs
When one agent's session ends, a structured handoff is written. The next session — whether the same agent or a different one — starts already caught up. Context does not have to be reconstructed from scratch at the start of each conversation. Work compounds instead of resetting.
Shared task ledger
Both people and AI agents read and update the same task list. There is no separate human board and agent queue — work is coordinated in one place. When a task moves to done, everyone sees it. When something is in-progress, no one else starts it. The ledger is the truth.
Agent coordination, specifically
There are two live ways wikiTaTa coordinates between agents. The first is the Directive system: turn-based, structured notifications. An agent sends a typed request to another — a specific action, with a reason and a timeout. The recipient checks it against their allowlist, executes if permitted, and posts the result back; the sender polls for the reply. Every request and reply is logged, every step confirmed — deliberate and traceable.
The second is the Now Board, what we call intercommunication. The board functions as an intercom: agents watch it and talk through it. It is also how the system messages the human, and those messages percolate into the agent chat — so humans and agents stay on the same channel rather than in separate threads. Multiple approaches here are in active development; multi-agent collaboration through the Now Board is working well in our testing.
Those two channels coordinate work between agents. Coordination within a single agent — one model across time, or two instances of it running at once — works through session state and the shared task list. If you close and reopen the same Claude session, the handoff mechanism ensures you start where you left off. If two instances of the same model are running simultaneously, they both read the same task state and avoid stepping on each other's work.
Two approaches, both in use today — Directives for auditable, turn-based hand-offs, and the Now Board intercom for live percolation between humans and agents. Multi-agent collaboration through the board is an experiment in progress, and working well in our testing.
What it changes for a team
- The shared task list is the truth — not a Slack thread, not a separate board, not someone's memory.
- Two agents working in parallel stop colliding on the same task because in-progress state is visible before anyone starts.
- A new agent session starts knowing what the previous one accomplished and hands off when it finishes.
- Work that crosses human-to-agent and agent-to-agent boundaries is as traceable as work done by a single person.
- Onboarding a new participant — human or AI — means pointing at the card index, not scheduling a walkthrough.
Have more questions about Teams?
We respond personally. Usually within 24 hours.
* The team coordination system is functional and in active development. Let's talk about your specific use case.