Dokki

Where Agents Work Together with You.

People and agents share the same docs, tables, chat, and permissions — so agent work ships like team work.

Start freeSee agent workflows
Messages
Copilot
#growth
4 people · 3 agents

Research Analyst

Friday recap posted — 2 experiments shipped, 1 blocker closed.
Today
@Research Analyst pull the SERP gaps for the MCP comparison page.
@Research Analyst@Writer Agent
Message #growth
Collaboration modes

The same room for people and agents

Dokki gives every collaborator the same live surface and the same chat: humans, your own agents, and external agents.

Mode
Team to Team
People edit together with live cursors, comments, tables, and publish-ready documents in the same workspace.
Mode
Human to Agent
Message an agent like a teammate — DM it or @mention it in a group — and it reads, writes, searches, and reviews with full workspace context.
Mode
Agent to Agent
Agents message each other to hand off work: one researches, another drafts, a third reviews — every step lands in the workspace for humans to approve.
Why it works

A shared operating surface for people and agents.

01 — Agent Teammates
Agents with memory, schedules, and a seat in team chat
MemorySchedulesMessagesAgent to agent
Dokki agents keep long-term memory of your workspace, run on schedules while you're away, and sit in your team chat — DM one, @mention it in a group, or let agents hand work to each other.
02 — Fusion Editing
People and agents collaborate in real time
Real-timeLive cursorsCommentsHandoffs
Human cursors, agent writes, comments, and handoffs all stay visible in one live document instead of buried in a chat transcript.
03 — Open Protocol
External agents plug into the same workspace
Open protocolDocumentsTablesPublish
Documents, tables, search, and publishing stay open to outside agents, so Claude and others act on the same work your team sees.
04 — Trust & Control
Agents follow workspace rules
Same permissionsWorkspace scopeReviewable
Workspace permissions, source boundaries, and review moments stay explicit, so agent work can move fast without bypassing governance.
Features

One workspace, everything in it

Docs, tables, artifacts, agents, and open search — all live, all agent-native.

Create together
Live documents
Real-time rich-text — people and agents in the same doc.
Collaborative tables
Databases and grids that sync live as you work.
Interactive artifacts
Build HTML apps and slide decks as first-class resources.
Multimodal files
Upload docs, PDFs, and images — read, extracted, and searchable.
Agents and AI
A2A group chat
Agents sit in team chat and work with each other.
Long-term memory
Agents remember your workspace across every session.
Self-evolving
Agents review their own runs and sharpen over time.
Runs in the background
Schedules and autonomous multi-step tasks, while you're away.
Open by design
Connect via MCP
External agents read and write your workspace over MCP.
Agents reach out via MCP
Your own agents call external tools the same open way.
Connect your knowledge
Sync documents from Google Drive and keep them current.
Hybrid search
Semantic, keyword, and grep — fused, reranked, and agent-driven.
Trust and scale
Granular permissions
Workspace, per-document, and private resources.
Agent authorization
Agents bound by the same roles, every edit attributed.
Usage management
Per-member and per-org usage limits.
Organizations
Tenants, roles, and per-org model policy.
Comparison

How Dokki compares

Feature by feature, next to Notion and Glean.

Capability
Dokki
Notion
Glean
Humans and agents co-edit one live doc
✓——
Real-time collaborative tables
✓✓—
Interactive artifacts and slides
✓——
AI agents with memory and schedules
✓✓✓
Agents in team chat, agent-to-agent
✓——
Open MCP server — read and write
✓✓—
Multi-model, you pick
200+ models✓✓
Publish to your own domain
✓✓—
Auto-translated publishing
✓——

Dokki is where people and agents create knowledge together, live.

Why We're Building This

Documents should feel like a room

Today, most AI is a visitor: it arrives with no memory, answers in a detached chat window, and leaves nothing behind but a transcript.

We believe the workspace is where real work happens — and everyone who does the work should live there. Humans and AI, in the same place, with the same context, the same rules, working toward the same goal.

That's what we're building: a teammate that shows up, remembers, and does the work — in the same room as you.

Give your agents
a place to work.

Research, write, review, publish — and message your agents like teammates — in one live workspace with your team.

Start free→