TL;DR: Install the Runbear plugin in Claude Code, run /runbear:deploy, then /runbear:connect-slack. Your local agent becomes an always-on Slack bot your whole team can use, no laptop required.
You can build a genuinely useful AI agent in Claude in an afternoon. Then you close your laptop, and it stops.
The fix is to deploy it: send your agent to Slack so it runs on its own, always-on, where your whole team can reach it without you in the loop.
| Dimension | On your laptop | Deployed to Slack |
| Availability | Runs only while your machine is awake | Always-on, 24/7 |
| Who can use it | Just you | Your whole team, in Slack |
| Identity | A local project | Its own Slack bot, with a name and icon |
| Permissions | Your access only | Each teammate authenticates as themselves |
| Your involvement | You run it for people | People use it directly, without you |
Here is how to do it.
How to deploy a Claude agent to Slack
You already have an agent running locally: a project with a CLAUDE.md, whatever skills it needs, and its integrations. Here is how to deploy it.
Prerequisites
- Claude Code installed, with the Runbear MCP server connected in your Claude Code session (the deploy runs through it).
- An existing local agent: a project folder with a CLAUDE.md, and optionally skills and MCP integrations.
Set it up once. Runbear's deploy tools ship as a Claude Code plugin (see the Agent Skills docs). Add the marketplace and install it one time:
/plugin marketplace add runbear-io/skills
/plugin install runbear@runbear-skillsThat gives you two commands: /runbear:deploy to deploy a project, and /runbear:connect-slack to give it a Slack bot. (The plugin lives in the runbear-io/skills repo.)
1. Run the deploy
From your project, run:
/runbear:deploy --new "My Agent"It packs your project and uploads it to a hosted agent (point it at an existing one by name instead of --new). Your files upload directly and securely, and secrets, .env files, and credentials are filtered out automatically, so they never leave your machine.
The deploy walks you through the rest right in Claude Code: it creates the agent, asks which integrations and skills to bring, packs the project, and offers to connect Slack.
2. Choose the integrations to bring
The deploy shows you the tools from your local setup (your MCP integrations) and asks which the hosted agent should carry. For each, you pick how your team authenticates:
- Per-user - each colleague signs in as themselves, so the agent respects their individual permissions (their Notion, their CRM access).
- Shared - everyone uses one org credential, for tools that have a single company account.
3. Pick the skills it ships with
Select which of your skills travel with the agent, from your project (.claude/skills/) and your user scope (~/.claude/skills/), so the hosted version behaves like your local one. The choice is saved in a .runbear/deploy.json manifest, so a redeploy brings the same set.
4. Connect it to Slack
The deploy offers to do this at the end, or run it any time:
/runbear:connect-slack "My Agent"Runbear gives your agent its own Slack bot, with its own name and icon, installed in your workspace. Now anyone can mention it in a channel or send it a direct message.
That is it. The agent you built on your laptop is now a shared teammate that runs 24/7, with no laptop required.
Agent Skills are now an open standard: what changes for a Slack agent
Short answer: nothing breaks, and your skills get more portable. As of October 2026, Agent Skills, the SKILL.md folder format Claude Code uses for reusable instructions, scripts, and reference files, is maintained by the Agentic AI Foundation under the Linux Foundation. Anthropic donated the name, logo, GitHub organization, and agentskills.io domain on October 22, 2026 (Anthropic's announcement).
Anthropic says more than 70 agent platforms, including ChatGPT, Cursor, GitHub Copilot, and VS Code, have integrated the format, and that 1 million skills have been published. Engineers from Anthropic, Google, OpenAI, and other companies maintain the standard as individuals, and the proposed roadmap covers versioned releases, a formal proposal process, and security guidance for people who write and install skills. The foundation added it alongside OpenAI's Open Responses, Hugging Face's OpenEnv, and Alibaba's OpenSandbox (AAIF announcement).
For a team agent, the practical effect is that a skill is a plain folder with a SKILL.md, not a format tied to one tool. When you run /runbear:deploy, you choose which skills from .claude/skills/ and ~/.claude/skills/ ship with the hosted Claude Agent SDK agent, and it loads them on its next activation when someone mentions it in Slack. If you would rather build the agent in the dashboard than in Claude Code, see the no-code Slack agent builder.
One habit worth keeping: a skill can include scripts, so read a third-party skill before you ship it to an agent your whole team can trigger. That is the gap the foundation's proposed security guidance is meant to close.
Hosting options for a Claude agent in Slack without your own server
Runbear hosts a Claude Code agent for you: /runbear:deploy uploads the project and /runbear:connect-slack gives it its own Slack bot, so you run no server and the agent keeps answering when your laptop is off. Runbear is one of five ways to put an AI agent in Slack. With Claude Tag (Anthropic's Slack app), Claude Code in Slack or Runbear you run no server, your own Slack app leaves the code and the hosting to you, and with Add to Slack it depends on the partner platform. If your company is on a Claude Team or Enterprise plan, Claude Tag runs in Anthropic's sandbox unless your organization sets up its own environment.
| Route | What you get in Slack | What you manage |
| Claude Tag (Anthropic) | @Claude as your organization's shared agent, which can load the CLAUDE.md files and skills from a repository an admin has granted to the channel | No server to run. It needs a Claude Team or Enterprise plan, and an admin sets each channel's connections |
| Claude Code in Slack (Anthropic's earlier app) | Coding requests to @Claude run as Claude Code cloud sessions on your own Claude account and connected repositories | A Pro or Max plan and a GitHub repository connected at claude.ai/code |
| Your own Slack app on a host such as Vercel | Whatever you build | The Slack app, its permissions and event handling, and its deployment to that host |
| Add to Slack from a partner platform | An agent you build in a partner tool such as n8n or Lovable | The agent, inside that tool |
| Runbear | Your Claude Code project as its own named agent your team shares, or an agent you create without code, on hosted infrastructure | Running /runbear:deploy and connecting Slack, or creating the agent in Runbear |
Claude Tag and Claude Code in Slack
Claude Tag loads the CLAUDE.md files and skills from a repository an admin has granted to the channel. In a channel, it works through the connections an admin set for that channel, plus your own claude.ai connectors once you allow them, and the MCP servers you configured on your own machine stay behind. It is Anthropic's current Slack product for Team and Enterprise plans. The earlier Claude Code in Slack app turns coding requests into cloud sessions on your own Claude account and connected repositories. It answers channel mentions only from people on Pro or Max plans, and only in workspaces not connected to Claude Tag (Claude Tag for Claude Code users, Claude Code in Slack).
Your own Slack app
You get full control and you maintain all of it. You create the Slack app, set its permissions and event subscriptions, verify that each request comes from Slack, and host it somewhere that can answer within Slack's three-second window. Vercel's Slack agent skill walks a coding agent through those steps and deploys to Vercel (Vercel's guide). From then on, the code and the hosting are yours to run.
Add to Slack
Slack launched Add to Slack in August 2026 with Hyperagent, LangChain, Lovable, n8n, NanoClaw, OpenAI, Runlayer, Skydive, Superhuman and Vercel (Slack's announcement). Runbear is not on that list, so this route fits an agent you already built in one of those tools.
Runbear
To deploy an agent you built in Claude Code to Slack for your team, install the Runbear plugin and connect the Runbear MCP server, run /runbear:deploy to upload the project, then run /runbear:connect-slack to give it its own Slack bot with its own name and icon. The deploy leaves your local MCP configuration file behind, so you pick the integrations to bring as part of the deploy. If you have not built an agent yet, you can create one in Runbear without code: see How to Add an AI Agent to a Slack Channel.
Why deploy a Claude agent instead of running it locally?
Deploying an agent is not just copying it to a server. It is the moment your private tool becomes something your whole team relies on. Three things change: it goes always-on, it gets its own identity in Slack, and, when it touches real tools, it respects who is asking.
We think the future of agents at work is not one giant assistant that does everything. It is dozens of small, purpose-built agents, each built by the person who understands the problem best, each deployed to the team that needs it. The person who lives in your support queue should build the support agent. The person who owns onboarding should build the onboarding agent. Claude makes that building step accessible to anyone. Runbear makes the last step, deploying it, just as easy.
This is already happening. At InMobi, the sales ops team built their own agent and deployed it to Slack, so the whole team uses it directly instead of routing every request through the person who built it.
Build is solved. Release is the part we are making easy.
FAQ
Does my computer need to stay on for the agent to run?
No. Once you deploy the agent to Slack with Runbear, it runs on hosted infrastructure, not your laptop. It keeps answering, posting, and reacting around the clock even when your machine is off.
How is this different from Claude's native Slack integration?
Claude's native option is Claude Tag, Anthropic's Slack app. It gives your organization one shared @Claude, which can load your repository's CLAUDE.md and skills and, in a channel, works through the connections an admin set for it, plus your own claude.ai connectors once you allow them. Deploying your own agent with Runbear ships the project you built, with the integrations you choose to bring, as a separate bot with its own name in Slack. If your organization already uses Claude Team or Enterprise and wants one shared @Claude, Claude Tag fits. If you want the project you built as its own named agent, deploy it with Runbear.
Are my code or secrets uploaded when I deploy?
Your project files are uploaded so the hosted agent has the same context, but secrets, .env files, and credentials are filtered out automatically before upload and never leave your machine.
Do teammates use their own tool permissions?
Yes. For per-user integrations, each teammate authenticates as themselves, so the shared agent respects each person's own access. Shared integrations use a single org credential instead.
Is it free to deploy a Claude agent to Slack?
Yes, to start. Runbear is freemium: you can deploy your agent to Slack on the free plan and upgrade as your team's usage grows.
Can I deploy the agent to other channels besides Slack?
Yes. The deploy commands connect Slack, and a Claude Agent SDK agent can also be added to Microsoft Teams from the agent's Chatbots tab in the Runbear dashboard (Microsoft Teams setup).
Do the skills I deploy work in other agent tools?
The skills themselves follow the open Agent Skills format, so the same SKILL.md folder can be loaded by other platforms that support the standard. On Runbear, they run on Claude Agent SDK agents.
Once deployed: a teammate asks the agent a question in Slack, and it answers with its sources and drafts a reply, always-on.
Build your agent in Claude. Release it to Slack with Runbear.
