From reporting to seeing
Status comes from the work in the room. Teams share changes and blockers without preparing another status readout.
Roomcord | AI decision engine
Know what needs your attention. Without chasing updates.
Roomcord brings team conversations, meetings and AI teammates into shared project rooms. We’re building its decision engine to surface decisions, owners, deadlines and blockers, with people confirming commitments.
Product screenshot
See the room your team works in
A shared chart, a teammate’s request and an AI reply thread sit in the same project conversation.
What is Roomcord and who is it for?
Roomcord is an AI decision engine for project teams. It brings people, conversations, meetings and files into one project room. We are building its decision engine so AI proposes source-linked decisions, owners and deadlines, people confirm them, and approvals and follow-up keep the same board current. It is for project owners, operations leaders and executives whose work stalls between discussion and delivery.
From discussion to decision
A request includes options, evidence, a recommendation and a decision deadline. Approvers settle what happens next.
From promises to a record
An accepted commitment has one owner, one date and a source. Changes retain who changed what and why.
Product screenshot
Find the original conversation
Search the room for a term and return to the matching messages. On mobile, this capture shows global search.
Follow one decision from conversation to closure
Illustrative product-launch workflow, not a customer result or a recording of the current interface. The intended outcome is an approved launch decision with an accepted owner, a visible first step and a traceable completion record.
- Conversation: the team discusses moving a launch to November 14 and attaches the readiness review.
- Suggested card: AI proposes the launch decision and its source. An unclear owner or date stays unresolved until someone supplies it.
- Confirmation: the project owner checks the proposed card against the discussion and corrects it. The change is attributed.
- Approval: the named approver reviews options and a recommendation, then approves, changes or reassigns by the ballot deadline.
- Commitment: Claire explicitly accepts delivery by November 14 and records the first step. AI does not silently assign the work.
- Follow-up: reminders and early slip reports update the same board. An unresolved blocker reaches the appropriate approver.
- Closure: the owner records the outcome with evidence. The ledger retains the decision, its reasons and later changes.
Run one meeting-to-follow-up cycle
Start with a practical method, then compare the tools for your actual handoff. The guide separates human-maintained records available today from decision-engine automation in development.
In development
From conversation to a live board
The next chapter of Roomcord builds on shared rooms, meeting capture and room chat. Auto-board, approval ballots, commit and chase, boss view and decision ledger are the decision-engine direction now in development; these sections describe the experience we are building.
Auto-board
Decisions, Approvals, Tasks, Milestones and Risks, extracted from room conversations and sources. Each card keeps one owner, one date and a link to the evidence.
Approval ballots
Bring a question, options and a recommendation. Approvers can approve, change or reassign against a clear deadline, with protected decisions requiring explicit approval.
Commit and chase
Owners confirm “I commit by this date.” Roomcord is being built to handle reminders, surface slipping dates and escalate unresolved blockers.
Boss view
See what is red, what is waiting on your decision and which commitments need attention across rooms.
Decision ledger
Keep who decided, when and why, the alternatives considered and links back to the original discussion.
Room agent
An AI teammate prepares drafts, gathers missing information and helps carry agreed work forward using the room’s context.
Meet to decide. Let the room carry the follow-up.
Our meeting direction connects the work before, during and after a call: a pre-read and decision ballot, live capture of owners and dates, then a source-linked recap and commitment requests. Approve an item asynchronously and remove it from the live agenda.
Before
Review what changed, red items and recommended decisions. Resolve the ballot early and give the team its meeting time back.
During
Focus on decisions: approve, change or reassign. Capture each agreement with its owner and deadline.
After
Confirm commitments and let reminders and escalation keep the next steps visible.
AI proposes. A human confirms.
An extracted card stays suggested until a person confirms it. Ownership is explicit, edits are attributed and sources stay attached.
How Roomcord works
Chapter 01
Every decision keeps its record.
Room context preserves the discussion, sources and responsible people. We are building the next step: a decision ledger and live cards that connect each agreement to its owner and date.
Chapter 02
Know what needs your decision.
Ask your personal Assistant about the rooms you can access. Our boss-view direction brings red items, pending approvals and commitments into one overview, with links back to the room.
Chapter 03
Bring specialist agents into the room.
Use configured agent templates or connect compatible external agents through A2A. Give each specialist a defined role and approved tools; a public marketplace with open discovery and billing is a future vision.
Chapter 04
One shared context for people and agents.
RoomAgent keeps room knowledge available while specialists help with research, writing and analysis. The shared conversation and sources give everyone a common basis for the next decision.
Chapter 05
AI proposes. People confirm.
Agents can compare options and prepare recommendations. In the decision-engine direction, extracted cards stay suggested until a person confirms the agreement, owner and deadline.
Chapter 06
Keep follow-up moving between meetings.
Use supported agent schedules for recurring work and summaries. We are developing commitment reminders and escalation so slipping dates become visible before the next meeting.
Chapter 07
Phone, desktop, meeting, one project room.
Keep conversations, files and meeting context together as you move between devices. The room connects the discussion to the work, so the next person can pick up the thread.
One room for people, AI, and the work between them
Roomcord is not another private chatbot window. It is a shared work surface where the team and its AI coworkers use the same conversation, sources, decisions, meetings, files, and current context to move work forward.
People and agents share one room
Teammates and AI agents work in the same conversation, not in separate bot windows. Everyone, human or agent, sees the same discussion.
A living Room Context
Each room maintains an auto-refreshed context document: project status, key decisions, and sources, kept current as the conversation moves.
Structure instead of scrollback
Instead of digging through message history, the room’s knowledge is organized into overview, decisions, and open topics you can read in one pass.
Context is what makes agents useful
Agents answer from what the room has actually discussed and decided, with grounded, specific responses instead of generic AI output.
Connect your computer to Roomcord
Pair a local computer and choose the project folder to connect. See the CLI setup guide.
Connect your computer with the CLI
Run these commands on the computer you want to pair. You need Node.js 22 or later and npm. To run coding tasks, install and sign in to Codex or Claude Code on that computer.
1. Install the CLI
npm i -g @fulldivevr/workcord-cli
2. Pair your computer
workcord machine pair
Open the displayed link or scan the QR code, then confirm pairing in Roomcord. Wait for the CLI to confirm that your computer is paired.
3. Connect a project folder
Replace /path/to/your/project with your actual project folder:
workcord machine start --root /path/to/your/project
Keep this terminal running while you use the connection. Press Ctrl+C to stop it. The --root option selects the working folder; it does not sandbox the coding agent. Tasks run with your local user’s access and without per-task approval.
Optional: run in the background
On macOS or Linux, use this instead of the foreground command:
workcord machine install-service --root /path/to/your/project
To stop and remove the background service:
workcord machine uninstall-service
Watch room memory in action
Room memory 0:26
A room answers questions from its decisions and history.
Illustrative examples feature fictional AcmeFlow and Ridgeway Logistics. Narration is synthetic (Victor). English audio; captions are not available.
What agents do in a room
Agents in Roomcord are room participants: mention them, give them a goal, or let them keep the room's knowledge current.
Agents
On-demand status reports
Ask @PM Buddy for a project status and get a formatted done / in progress / blocked report built from the room.
Agents
Multi-agent discussions
Give several agents a goal and let them debate it , the discussion ends with an auto-generated conclusion card.
Shadow
Cross-room answers
Your personal agent Shadow can check the rooms you have access to and report back , launch dates, risks, decisions.
Controls
Per-agent behavior settings
Decide when each agent responds , mention only, proactively, or suggestions , and whether it replies in a thread or in chat.
Voice
Voice sessions with agents
Talk to an agent in a full-screen voice session and get live answers drawn from your rooms’ context.
Context
Room Context in chat
A summary card surfaces the room’s current context and topics right in the conversation, one tap from the full document.
Answers across your rooms
Personal assistant 0:24
One assistant answers across rooms the user can access.
Illustrative examples feature fictional AcmeFlow and Ridgeway Logistics. Narration is synthetic (Victor). English audio; captions are not available.
Coordination without ceremony
Use Roomcord when a stream of work needs shared memory, not another place to copy updates. Threads, pins, reactions, mentions, and meetings all feed the room context so the team and its agents keep the same operating picture.
Threads and quoted replies
Keep side discussions attached to the message they answer, so decisions stay traceable.
Pins and reactions as prioritization
Pin an @all list and let the team vote with reactions , lightweight prioritization without a separate tool.
Mentions for people and agents
@mention a teammate, an agent, or @all. Agents respond as participants, in place, where the question was asked.
Meeting Mode
Long-press the mic to record a meeting , up to two hours. Roomcord transcribes it and Shadow posts a clean recap with a shareable permalink.
Operational outcome
Less status-chasing and fewer recap meetings: decisions, priorities, and meeting notes are captured where the work already happens.
Multi-agent coordination across compatible systems
Roomcord is the collaboration surface, not a replacement runtime for every agent. Bring A2A-compatible external agents into the room alongside Bewize-managed agents. Bewize Hub separately hosts and governs agents that run on its managed infrastructure.
Hosted and governed agents
Agents managed by Bewize Hub run with isolated runtimes, policies, and usage visibility. Roomcord is where people work with them and compatible external agents.
A2A identity per room
Rooms expose agent card URLs and API keys, so any A2A-compatible agent can join a room with a managed identity.
Inspectable tool calls
When an agent uses a tool, the call renders as a collapsible detail in chat , you can always see what an agent actually did.
Agent flavors
Internal one-click agent templates spawn a pre-configured agent for a room , a PM agent, a QA agent , without implying an open third-party marketplace.
From specialist to room participant
Agent lifecycle 0:37
Create a specialist, add it to a room, and see it use room context.
Illustrative examples feature fictional AcmeFlow and Ridgeway Logistics. Narration is synthetic (Victor). English audio; captions are not available.
Bring agents built with your framework
Roomcord supports external agents exposed through A2A. These frameworks document ways to serve A2A agents; configure and authenticate an A2A endpoint before adding it to a room. This is protocol compatibility, not a claim of preconfigured or certified vendor integrations.
Google ADK
Google Agent Development Kit
Expose an ADK agent through A2A, then bring its endpoint into a room.
A2A setup docs →
Microsoft
Microsoft Agent Framework
Host a framework agent as an A2A service for room participation.
A2A setup docs →
Agno
Agno AgentOS
Expose Agno agents, teams, or workflows through its A2A interface.
A2A setup docs →
A2A deployment path
LangChain / LangGraph
Serve a LangChain agent over A2A using a documented deployment template; ordinary LangGraph agents do not expose A2A automatically.
A2A setup docs →
AWS
Strands Agents on AgentCore
Deploy a Strands A2A server on Bedrock AgentCore and connect the configured endpoint.
A2A setup docs →Explore the decision engine
Follow each part of the same workflow in more detail.
Auto-board
How conversations become five types of source-linked cards.
Learn more →Approval ballots
How recommendations reach a decision with approvers and deadlines.
Learn more →Commit and chase
How accepted commitments receive reminders and contextual escalation.
Learn more →Decision meetings
How pre-reads, ballots and recaps replace status readouts.
Learn more →Boss view
How leaders see exceptions, pending decisions and delivery across rooms.
Learn more →Decision ledger
How decisions keep their reasons, sources and attributable history.
Learn more →Room agent
How an AI teammate prepares drafts and gathers missing information.
Learn more →What can your team use today?
As of October 21, 2026, shared rooms, chat, files and meeting capture form the existing Roomcord foundation. Auto-board, approval ballots, automated commitment follow-up, boss view and decision ledger are in development. This page announces the intended experience; pilot scope and availability are agreed for each team. Existing room-memory demos below illustrate the foundation, not completion of the full decision loop.
Keep people in control
Suggested cards require human confirmation. Money above a room’s limit, legal matters and people decisions are protected items that require explicit approval. Any timeout policy must retain those exceptions.
Start with one recurring decision
Bring one project, a recurring approval or meeting, the decision owner and relevant sources. Map the first workflow and measure decisions closed on time, meeting time and manual follow-up before and during use.
Explore current Roomcord capabilities
AI on your computer
Pair a computer with Roomcord, connect a project folder and run local coding tasks through Codex or Claude Code. Keep the result connected to your project room.
Learn more →Interactive answers and dashboards
Bring charts, tables, timelines, diagrams and supported interactive dashboards into the room where your team discusses the work.
Learn more →Work with several AI specialists in one conversation
Bring relevant agents into a room discussion and keep their replies visible to the team. Progress and stage indicators help people follow ongoing work; improved thread replies keep a response attached to its question. Cross-room work uses the access available to the participating agents. Add specialists to prepare options and evidence, while people retain responsibility for decisions and accepted commitments.Build a workflow inside your room
Create mini-apps that keep their data in the room and let members and AI agents call their functions. Start with the starter, poll or approvals example, test locally, and have a room admin approve the app. Build mini-apps for Roomcord.Common questions
What problem does Roomcord solve?
Every project gets a room. We’re developing Roomcord to turn its conversations, meetings and files into a live board of decisions, owners and deadlines, with approvals and follow-up built in. Your team does the work. You see progress without chasing reports.
Is Roomcord a chat clone?
Roomcord is our main product: an AI decision engine built around project rooms. Shared chat, meetings and files provide the foundation; auto-board, approval ballots, commitments and follow-up are the direction in development. Bewize Hub and Wize Browser are supporting components for agent execution and web work.
Where do the agents run?
Bewize-managed agents run on Bewize Hub, its agent runtime control plane. Compatible external agents can run in their own environments and join Roomcord rooms through A2A; Roomcord does not control their underlying runtimes.
How do agents avoid making things up?
Room context and inspectable tool calls help people check an agent’s answer, but they do not guarantee accuracy. Verify consequential claims against the linked source. In the developing decision engine, AI-created cards stay suggested until a person confirms them.
Who controls how agents behave?
You do, per room and per agent: whether an agent responds only on @mention, proactively, or as suggestions; whether it replies in a thread or directly in chat; and you can remove an agent from a room at any time.
What is Shadow?
Shadow is the personal AI agent every Roomcord user gets. It has access to the same rooms you do, answers questions across them, drafts replies when @mentioned, joins voice sessions, and turns recorded meetings into shareable recaps.
Can we connect our own agents?
Yes , Roomcord speaks A2A. Any A2A-compatible agent can join a room using the room’s agent card URL and an API key you issue and can revoke.
Start your first decision room
Bring one recurring meeting or approval workflow. Explore how Roomcord can connect the conversation, decision, owner and deadline, and which decision-engine capabilities are available for your pilot.
Architecture conversation
Tell us what your team needs to control
Share your deployment boundary, number of agents, work surfaces, and governance requirements. We will reply by email to arrange a focused technical discussion.
Email the Bewize teamThis opens your email application. Read our Privacy Policy.