Work with AI the way you work with people.
A whiteboard session beats an email thread. A room with AI beats a prompt box. Same principle, higher bandwidth.
Higher fidelity than chat
No matter how good you are at prompting or context engineering, a chat window is a bottleneck. A meeting is higher bandwidth — the AI can show you its work, you can react naturally, and the whole loop is tighter.
Multiplayer by default
Invite your team into the same room as the agent. Review a financial model together. Get a presentation delivered to five people at once, not copy-pasted into five separate threads.
Zero friction
Click a link and you’re in. No prompts to write, no files to attach, no context windows to manage. The AI already has the room, the screen, and the tools. You just show up.
Explore two strategies. Merge the best.
Fork a meeting into parallel tracks. Each track gets its own agents, its own context, its own deliverables. When one path proves stronger, merge it back. The other track’s work is preserved — nothing is lost.
Fork
Branch at any decision point.
Disagree on a strategy? Fork the meeting. Each track runs independently with full agent support. No context duplication — the fork carries the full semantic history.
Parallel execution
Different teams, different approaches, same clock.
Both tracks produce deliverables simultaneously. The financial analyst runs numbers in Track A while the strategist builds a narrative in Track B.
Join
Semantic merge, not copy-paste.
When you merge, the system reconciles decisions, surfaces conflicts, and produces a unified outcome. Like git merge, but for meetings.
Three steps. Real outcomes.
Pick an agent — or don’t
Start with a preset — financial analyst, researcher, strategist — or start from scratch. A blank Modyl agent learns your style, adapts to your domain, and improves with every session. No preset required.
Start a room
Create a room, invite your team, and the agent joins alongside you. Camera, screen, voice, interactive surfaces — everyone in the same space.
Ship the outcome
Walk out with deliverables, not action items. Documents, models, presentations, and follow-ups — produced during the meeting, not after it.
Coming soon: Fork a meeting into parallel tracks. Explore two strategies simultaneously and merge the best outcome.
Head starts, not hard requirements.
Presets give you a running start with tailored tools and templates. Or skip them entirely — a vanilla Modyl agent adapts to your domain and improves itself with every meeting.
Financial Analyst
Three-statement models, LBO analysis, comparable company analysis. Excel-grade spreadsheets generated in-meeting.
Deep Researcher
Multi-source investigation with citation tracking. Synthesizes findings into structured reports with provenance.
Business Strategist
Market analysis, competitive intelligence, strategic frameworks. Delivers presentation-ready decks.
Optimization Solver
Constraint optimization, scenario modeling, sensitivity analysis. Turns meeting discussions into quantitative models.
Business Analyst
Data pipeline design, visualization, and automated reporting. Turns raw data into actionable dashboards.
Scientist
Hypothesis testing, experiment design, statistical analysis. Rigorous methodology in conversational form.
These are just the starting points. Search above for more presets — or start with a blank agent that adapts to you.
Show your work. See their reaction.
Tighter feedback loops produce better outcomes. Install the skill or the SDK. Create a room. Send a link. The rest happens in real time.
Install the skill
Works with Claude Code, Codex, Gemini CLI, and any agent that supports the agentskills.io standard.
Install the SDK
For agents built on custom frameworks, LangChain, or direct Python. Full control over the room lifecycle.
# The agent runs this. The user gets a meeting link.
from modyl import Agent, Room
agent = Agent("my-agent", capabilities=["a2ui", "camera"])
room = await Room.create("review-session")
# User clicks this link to join.
join_url = room.get_join_url(role="human")
# Agent joins and presents its work.
await agent.join(room)
await agent.push_surface(analysis_results)Present
Show the work, not just a summary of it.
Push live charts, data grids, and interactive surfaces directly to every participant.
Observe
Close the feedback loop.
Camera, voice, and screen sharing close the loop. The agent processes reactions and iterates on its output in real time.
Retain
Nothing lost between sessions.
Every correction, approval, and decision feeds into persistent memory. The next session resumes where this one ended.
Build in minutes.
A Python SDK, an extension system, and a protocol. Everything you need to give your agents presence.
surface = A2UI.surface("revenue-chart", components=[
A2UI.heading("Q4 Revenue Breakdown"),
A2UI.live_chart(data=revenue_data, type="bar"),
A2UI.data_grid(rows=line_items, editable=True),
A2UI.button("Export to Excel", action="export_xlsx"),
])
await agent.push(surface)First-class infrastructure.
Rooms aren’t just meetings — they’re compute primitives. The room is the hard drive.
Rooms as Primitives
A room is a mutex, a shared state space, a collaboration surface. Your agent gets a persistent workspace with every join.
A2UI Protocol
Push declarative JSON UIs to any connected human. Data grids, live charts, forms, markdown — rendered natively across platforms.
Capability Routing
Discover and invoke device capabilities on connected clients. Calendar, contacts, camera, location — through a typed command protocol.
Memory That Persists
Session context, relational knowledge, and meeting history survive across rooms. Your agent remembers everything.