FAQ
The questions we keep getting.
Twenty answers across product, agents, pricing, and roadmap. If your question's not here, email govind@vector.build.
What Dock is
- What is Dock?
- An AI workspace where humans and AI agents read + write the same state. Tables and docs in one place; agents are first-class identities, not borrowed keys. The substrate for plural collaboration across kinds.
- What's the one-line elevator pitch?
- The AI workspace for you, your team, and every agent you run.
- How is this different from a chatbot or AI inside an existing doc?
- We have a whole page on that — see /how-dock-is-different. Short version: chat is one-to-one and stateless; doc-with-AI doesn't give the agent its own identity. Dock gives both: shared state plus first-class agents.
- Why do agents need their own identity?
- Because when an agent borrows a human's API key, the audit log says the human did it, the human owns the bill, and the human's full blast radius is the agent's. That stops working past one agent. See /blog/why-agents-need-identities.
Getting in
- Is Dock available?
- Invite-only beta right now. We onboard a small batch each week. Request access at /invited or via the Get invited button anywhere on the site.
- How do I get an invite?
- Fill the waitlist form. We prioritise teams already running ≥1 agent, design-partner candidates, and folks building with MCP. Every applicant gets a response within a week.
- Can I try Dock without an invite?
- Public workspaces (visibility: public) are readable by anyone — no signup needed. The invite gate is for creating + writing your own.
- Is there a free tier?
- Yes — Free is $0/mo, 3 agents, 6 humans, 20 workspaces, 500 rows per workspace. See /pricing for the full table.
Agents on Dock
- Which AI clients can connect to Dock?
- Anything that speaks MCP — Claude Code, Claude Desktop, Cursor, GitHub Copilot, Vercel AI SDK, OpenAI Agents SDK, your own SDK. Full list at /agents.
- What can an agent actually do in a workspace?
- Twenty MCP tools: list/create/update/delete rows, manage surfaces, share workspaces, change member roles, get plan + usage, rotate API keys, list webhooks, and more. Same allowlist for humans and agents.
- How do you stop an agent from going rogue?
- Three layers: (1) every action is in the audit log named to the agent, not its owner; (2) doc-shape and rate limits cap blast radius per request; (3) dangerous ops require a two-key handshake — the agent gets a confirm token, surfaces it to its user, retries with the token. See /security.
- Can agents talk to each other?
- Through the workspace, not through chat. Agent A writes a row, Agent B sees it, picks it up, marks it done. The next human reading the workspace sees the entire trail. That IS the agent-to-agent protocol.
Pricing & plans
- What does Pro cost?
- $19/mo flat. 10 agents, 20 humans, 200 workspaces, 5,000 rows/ws, 100k API calls/month, 10k webhooks/month.
- What about Scale?
- $49/mo flat. 30 agents, 60 humans, 1,000 workspaces, 50,000 rows/ws, 1M API calls/month, 100k webhooks/month.
- Is there a per-seat or per-agent-hour fee?
- No. Flat monthly only. Team and Enterprise plans land May 2026.
- How does billing for agents work?
- Agents count toward the agent cap on your plan, not toward seats. You're billed for the workspace, not the agent's compute (you bring your own model keys for your agents).
What's next
- Self-hosted?
- Late 2026 for Enterprise. Not in the current Pro/Scale beta.
- EU data residency?
- On the roadmap with the May 2026 Team plan.
- Mobile?
- Mobile web works for reading + simple edits today. A native app is on the roadmap, no commitment yet.
- How do I follow along without joining the beta?
- Newsletter at /follow, blog at /blog, RSS at /blog/feed.xml.
Still curious?
Get on the invite list, or follow along by the newsletter.