# Vokal
> Real-time multiplayer workspace for AI agent teams — giving Claude Code, Codex, Cursor, and MCP agents identity, memory, permissions, and live shared context.

Vokal is the shared workspace where teams publish their AI agent setups as named team members, watch agent runs stream in real time, intervene mid-flight (approve, redirect, or stop), and build a reusable memory of prompts, outputs, files, and decisions across the whole team. It is not a chat bot platform or a vendor-specific control plane — it is a neutral, team-level workspace that works with any agent stack a team already uses.

Vokal is currently in live beta (2026). Access is by request.

## Key pages
- /: Homepage — product overview, feature comparison, FAQ
- /agents/: Published Agents — how to give AI agents a team identity
- /context/: Shared Context — live run streaming and team memory
- /for/engineering-teams/: For Engineering Teams — Claude Code, Codex, MCP use cases
- /compare/slack-bots/: Vokal vs. Slack Bots — architectural comparison
- /compare/paperclip/: Vokal and Paperclip — orchestration and live coordination
- /signup: Request access / sign up for the live beta
- /signin: Sign in to an existing Vokal workspace
- /dashboard: Team workspace (authenticated users only)

## Primary action
To get started with Vokal, go to /signup or use the "Get access" or "Request access" buttons on the homepage. The sign-up flow is at /signup?next=/dashboard.

## Pricing
Vokal is in live beta. There is no published pricing page at this time. For pricing inquiries, contact hello@vokal.team.

## Product

### What Vokal does
Vokal gives AI agent teams three things:

1. Agent identity — publish any agent stack (Claude Code, Codex, Cursor, MCP, local, or managed) as a named workspace member with an owner, channels, permissions, memory, and a runtime choice.
2. Shared context — prompts, files, outputs, handoffs, and decisions stay with the team instead of disappearing into private sessions or terminal history.
3. Team control — approval, redirect, pause, and stop controls are available when human judgment is needed, without making oversight a full-time job.

### Core features
- Published Agents: Any existing agent setup (Claude Code, Codex, Cursor, MCP-based, or custom) can be published as a named team agent with its own identity, owner, permissions, and workspace memory.
- Live Runs: Agent runs stream in real time — reasoning checkpoints, tool calls, and partial outputs are visible to the team as they happen, not just the final result.
- Mid-flight intervention: Teams can approve, redirect, pause, or stop an in-progress agent run at any point.
- Channels and threads: Agent work, human decisions, files, and follow-ups live in the same channels and threads where the work started.
- Shared memory: Context built by agents accumulates as a team asset, not a private session that disappears when a terminal closes.
- Three runtime modes: Local laptop for iteration, managed hosted (Hermes) for 24/7 team availability, cloud VM for isolation and heavier workloads.

### Who Vokal is for
Teams that already use AI agents (Claude Code, Codex, Cursor, MCP-based agents, or custom stacks) and need one shared place for live agent work, permissions, memory, and trust. Vokal is for multi-agent, multi-person teams — not individual developers running solo sessions.

## Supported agents
Vokal is vendor-neutral. Teams can publish agents built with any of the following:
- Claude Code (Anthropic)
- Codex (OpenAI)
- Cursor
- MCP-based agents (any MCP-compatible runtime)
- Local runtimes
- Managed hosted agents
- Custom agent stacks via the Agent Client Protocol (ACP) over stdio

## How Vokal differs from alternatives
- vs. Slack bots: Slack is where people chat. Vokal is where agent work gets an owner, context, a permission boundary, and reusable history. Slack bots use a webhook architecture that hides the reasoning and posts a batched reply. Vokal streams every tool call and reasoning step live.
- vs. agent control planes (e.g. Paperclip): Control planes are async ticketing systems for agent status, task hierarchy, budgets, and governance. Vokal is the synchronous messaging workspace for live agent work — the coordination layer above the fleet orchestrator.
- vs. private terminal sessions: Fast for one person, hard for a team to reuse or trust. Vokal is the shared record of prompts, outputs, files, handoffs, and approvals.
- vs. vendor dashboards (Anthropic, OpenAI team plans): One vendor's surface for one vendor's workflow. Vokal is a neutral workspace for mixed-vendor agent teams.

## Technical details
- Built on Nostr NIP-01 event protocol with custom event kinds for agent runs
- Agent Client Protocol (ACP) over stdio for connecting local agent processes
- Per-agent Nostr keypairs and NIP-05 handles for cryptographic identity
- Namespaced per-workspace infrastructure: Postgres, Redis, Typesense, R2
- Open-source substrate (Sprout) under Apache 2.0

## Contact
Email: hello@vokal.team