AICoven 2.0
The private workspace layer for AI agents.
Models change. Agents change. AICoven keeps your projects, decisions, roles, memory, and policies together as one durable context.
Not another disposable chat thread. A durable workspace for serious AI work on iPhone, iPad, and Mac — with your own provider keys and your context under your control.
Models change. Agents change. Context should not.

What’s new
AICoven 2.0 turns chat into a workspace layer.
The update is bigger than another model picker. It is a private continuity layer for humans, roles, provider accounts, memory, external agents, and native Apple devices.
“Models change. Agents change. Context should not.”
Why AICoven
Your thread is the workstream, not the model session.
Serious AI work does not fit inside one vendor chat. AICoven gives the work a durable home, then lets the right role, model, agent, or local tool lane pick it up without making you start over.
Disposable chat
- · Context scattered across ChatGPT, Claude, IDEs, and notes
- · Every model swap starts with another handoff summary
- · Decisions buried in scrollback instead of workspace state
- · Memory tied to a product account you do not control
AICoven workspace
- · One durable context for each project or domain
- · Role-based agents that share approved Knowledge Hub memory
- · BYOK routing across cloud providers and user-hosted agents
- · Continuity across iPhone, iPad, Mac, and MCP-connected tools
How the work stays alive
One workspace layer above many models and agents.
AICoven does not replace every model or runtime. It gives them a shared place to draw from, return to, and hand work through.
- Workstream1
The thread holds the work
Project goals, decisions, open questions, and next steps stay attached to the workspace instead of disappearing into scrollback.
- Memory2
Knowledge Hub makes it accountable
Approved facts, policies, sources, and skills become shared context that roles can reuse and users can inspect.
- Routing3
The right executor picks it up
Use BYOK cloud models, Hermes, OpenClaw, or execution on your Mac via a connected local agent—with AICoven routing each turn and preserving workspace continuity—without moving the work to another product.
- Continuity4
Devices and tools stay in sync
Continue on iPhone, iPad, Mac, or an MCP-connected coding agent while AICoven remains the workspace layer.
Bring your own model, agent, and tools
{
"workspace": "AICoven Product Direction",
"roles": ["Designer", "Engineer", "Researcher"],
"routing": ["BYOK cloud", "Hermes", "OpenClaw", "Mac local tools"],
"context": ["thread", "Knowledge Hub", "policies", "sources"],
"mcp": "workspace memory for external agents"
}AICoven is the control surface. Cloud models, user-hosted agents, Mac-local execution via your connected agent—with AICoven routing each turn—and MCP-connected clients can work from the same private workspace context.
Built for work that outlives one chat.
Keep project context, decisions, sources, and next steps together instead of reconstructing work from old chats.
Store approved facts, policies, and sources so each role can use the same accountable context.
Connect your own provider accounts. Keys and content are encrypted at rest while AICoven orchestrates each turn.
Route work by role to cloud providers, Hermes, OpenClaw, or execution on your Mac via a connected local agent—with AICoven routing the turn and preserving workspace continuity—when the task needs your machine.
Let external agents and scripts query scoped workspace memory instead of copying project rules into every prompt.
Collaborate in private workspaces where each member keeps their own provider keys and tool connections.
AICoven Local — free and open source
Prefer a local lane? AICoven Local is a fully open-source client you run on your machine, inspect, and shape yourself. You host the agent process and bring your own API keys—this does not guarantee offline or on-device model inference.
Run the open-source client on your machine and inspect how it works. You host the agent process—not necessarily offline inference.
Contribute features, report bugs, and shape the roadmap in public.
Use your own provider accounts instead of buying resold inference.