Dust organizes AI for employees. Runbear works with everyone else too.
Dust is an internal AI workspace: Spaces, roles, and admin analytics for governing how your employees use AI. Runbear deploys named agents into Slack, Microsoft Teams, and Discord, including the shared channels where customers and outside vendors already work, no Runbear account required.
No credit card needed
Runbear is a Slack-native AI teammate. You create named agents, connect 2,000+ tools, and deploy them into the channels your team already works in. People outside your company can use those agents in a shared channel without signing up for anything.
Dust is an internal AI workspace. It organizes agents, company knowledge, and tools around workspace membership, with Spaces, roles, and admin analytics for governing how employees use AI.
Inside vs. shared
Dust serves your workspace.
Runbear serves the conversation.
Dust is genuinely strong at governing AI for employees. The question that changes your rollout is simpler than a feature count: is everyone who needs this agent on your payroll?
- Slack-first. Mention an agent where the work already happens. No dashboard round trip.
- Named agents. @Support, @Onboarding, @Billing, each with its own instructions, knowledge, and tools.
- Open to non-members. Customers and vendors in a shared channel use the agent without a Runbear account.
- Claude Code, deployed. Ship the CLAUDE.md, skills, and subagents you already built to a hosted agent.
- Workspace-first. Agents live in the Dust web app, organized by Space and role.
- Strong governance. Company, group, and personal scoping for knowledge and tools, with per-user and shared auth.
- Admin analytics. See which agents, tools, and people are actually being used.
- Members only. Using an agent means joining the Dust workspace first.
Side by side
The full feature matrix.
| Feature | Runbear | Dust |
|---|---|---|
| What it is | Slack-native AI teammate you deploy into channels | Internal AI workspace for employees |
| Where you use it | Slack, Microsoft Teams, Discord, and the web app | The Dust web app first; Slack and Teams as integrations |
| Who can use it | Anyone in the channel, including customers and vendors with no account | Members of your Dust workspace |
| Named agents | @Support, @Onboarding, @Billing mentioned by name right in the thread | Agents are named in the workspace; channel use runs through the integration |
| Integrations | 2,000+ via Pipedream, MCP, and native OAuth | A curated set of native connectors plus MCP |
| Triggers | Cron, event, channel, file, webhook, and external app events | Schedules, webhooks, and Slack Workflows |
| Chained automation | Event triggers fire an agent when work finishes in a connected tool | Harder to express; the trigger vocabulary is narrower |
| Claude Code projects | Deploy your local project, CLAUDE.md, skills and subagents included | Runs Claude models, but the local project itself is rebuilt by hand |
| Agent setup | Describe the job and the builder wires instructions, knowledge, and tools | Form-by-form configuration; templates exist but stay manual |
| Access control | Per-user OAuth, role-aware knowledge bases, RBAC, SOC 2 Type II | Spaces, groups, and roles, with clear warnings before you overshare |
| Usage analytics | Per-agent activity, credits, and usage reporting | Workspace analytics broken out by agent, tool, and person |
Where the work happens
The agent belongs in the channel, not behind a login.
Every AI tool claims a Slack integration. What matters is whether chat is the product or a side door, and that shows up in the first ten minutes of onboarding.
Install the Slack app, mention @Support, and the answer lands in the thread with its sources. The channel is the product surface, so adoption does not depend on convincing anyone to open a second tab.
Dust would rather host the conversation on its own dashboard. Setting up a workspace, we were never nudged toward Slack once during onboarding, and finding the Slack connect button took more than three minutes. That is a product center of gravity, not an oversight.
Who can use it
Your customers are not on your payroll.
Dust's access model assumes the person asking is an employee. That is the right call for internal knowledge work and the wrong one the moment the asker is a customer, a candidate, or an outside vendor.
Put an agent in a shared Slack channel and the customer on the other side simply uses it. No Runbear account, no seat, no invitation into your workspace. That is what makes candidate-facing bots, partner support channels, and vendor-facing agents possible at all.
Using a Dust agent means joining the Dust workspace. For an internal rollout that boundary is a feature. For a customer-facing workflow it is a wall with no route around it.
Bring your own agent
Ship the Claude Code project you already built.
Most teams already have a working agent. It just lives on one laptop, in one repo, driven from one terminal, which means exactly one person can use it.
Install the Runbear plugin in Claude Code and run /runbear:deploy. Your CLAUDE.md, skills, subagents, and docs ship to a hosted Claude Agent SDK agent that you connect to Slack. The setup you tuned locally becomes a teammate anyone can mention.
Dust runs Claude models, selectable per agent, and its CLI exposes Dust agents to Claude Code over MCP. That is the reverse of an import. The project you already built does not come with you. You re-express the instructions in the agent builder and recreate your Claude Skills as Dust Skills.
Choose the right tool
When to reach for each.
Choose Runbear if…
- Your team lives in Slack and will not adopt another dashboard
- Customers, candidates, or vendors need the agent without joining your workspace
- You want named agents people mention by name, like @Support and @Onboarding
- You already have a Claude Code project you would rather deploy than rebuild
- You need triggers that fire when work finishes somewhere else
- You want 2,000+ integrations instead of waiting for a native connector
Choose Dust if…
- Your first problem is governing employee AI use, not reaching outside users
- You need Space-level separation of knowledge and tools across company, group, and person
- Admin-facing workspace analytics are a primary requirement
- You want everyone working inside one AI web app rather than in chat
- Your workflows are self-contained question and answer rather than chained across tools
- You prefer a curated set of native connectors over a large marketplace
Questions you're probably asking
Runbear and Dust, side by side.
Put the agent where the conversation is.
10-minute setup. Free to start, no credit card needed. Bring a real workflow and we'll wire it live on a demo call.
No credit card needed