# Slack Agents

Give people an agent where they already talk. Each agent gets its own Slack app, its own instructions, tools and files, and a full record of every conversation in Everruns.

[← Back to use cases](/use-cases/)

![Several Everruns agents answering in Slack, each as its own app](/use-cases/slack-agents/stacked-chats.png)

## The Problem

Most teams already ask the same questions in Slack every day: how many vacation days carry over, whether the office is open on a holiday, who to ask about expenses. A person answers, or the question gets lost in a channel.

An agent can answer these, but a Slack bot that runs an agent is more work than the agent itself:

- **Hosting**: a public HTTPS endpoint that answers Slack within three seconds and verifies every request
- **The Slack app**: scopes, event subscriptions, a manifest, installation and tokens for each bot
- **Conversation state**: deciding which messages belong to which conversation, and keeping history between messages
- **Delivery**: retries when Slack or the model provider fails, and no duplicate replies
- **Control**: approving risky actions, seeing what the agent did, and stopping it

Most of this is the same for every agent you build.

## How Everruns Helps

In Everruns, Slack is one channel of an agent. You write the agent, Everruns does the rest.

1. **Define the agent**: instructions, model, capabilities, files and MCP tools. In the UI, as an `agent.toml` file, or by asking Codex or Claude Code to create it through the Everruns MCP server.
2. **Add a Slack channel**: choose one session per thread, per channel, or per user.
3. **Add to Slack**: Everruns creates the agent its own Slack app, with its name and avatar, and opens Slack's consent screen. Click Allow.
4. **Invite and talk**: `/invite @Company Info` in a channel, then mention it or send it a direct message.

An organization admin connects the Slack workspace to Everruns once. After that, each agent installs with **Add to Slack**.

## Example: Company Info

A company handbook agent for a team workspace. It answers questions about time off, holidays, expenses, equipment and offices from a handbook folder, names the file it used, and points to the right person when the handbook does not say. Each Slack thread is a separate session, so questions do not mix.

![Company Info answering handbook questions in a Slack thread](/use-cases/slack-agents/company-info-slack.png)

The whole agent is one folder:

```text
company-info/
├── agent.toml
├── instructions.md
└── handbook/
    ├── index.md
    ├── time-off.md
    ├── holidays-2026.md
    ├── expenses.md
    ├── equipment.md
    ├── offices.md
    └── people.md
```

`agent.toml` sets the model, the handbook files and the Slack channel:

```toml
schema_version = 1
name = "company-info"
display_name = "Company Info"
description = "Answers questions about Fernhill policies, holidays, offices and who to ask, from the company handbook."
capabilities = ["current_time", "session_file_system", "slack", "channel_context"]
files = ["handbook/**"]

[model]
provider = "openai"
model = "gpt-6.2-sol"

[channels.slack]
type = "slack"
[channels.slack.config]
session_strategy = "per_thread"
```

`instructions.md` is the agent's job description:

```markdown
You answer employee questions about Fernhill in Slack.

- Answer only from the files in handbook/. Read handbook/index.md first, then the file that covers the question.
- Name the file you used at the end of the answer, like: (handbook/time-off.md).
- If the handbook does not cover it, say so and name the person to ask from handbook/people.md. Do not guess.
- For dates, check today's date in America/Chicago first.
- Keep answers to two to four short lines. Use the asker's first name.
```

Import it with the CLI, or as a ZIP in **Agents > Import**:

```bash
everruns agents import ./company-info
```

Fernhill is a made-up company. Replace the handbook files with your own; a larger document set fits the [Knowledge](https://everruns.com/docs/capabilities/knowledge/) capability, which adds search over file collections.

## Example: Kids Duties

A chores and screen time agent. It knows the dish rotation and the screen time rules, checks the date, and remembers what was done today. It also posts a morning briefing to Slack on a schedule.

![Kids Duties in Everruns and the same agent posting its morning briefing in Slack](/use-cases/slack-agents/kids-duties-slack.webp)

We built a smaller version from Codex with one prompt, after adding the Everruns MCP server with `codex mcp add everruns --url https://<your-everruns>/mcp`:

```text
Create an Everruns agent for kid duties.

Name: kid-duties. Display name: Kid Duties.
Model: GPT-6.2 Sol. Capabilities: current time, Slack, channel context.

What it does:
- Dishes rotate one kid per day: Anna, Ben, Chris, Dana, then Anna again. Anna was on dishes on Monday 2026-10-26.
- Screen time is 1 hour on school days and 2 hours on Saturday and Sunday, and only after that day's duties are done.
- When a kid says a duty is done, confirm it and remember it for the rest of the day.
- Check today's date in America/Chicago before answering. Keep replies to one to three short lines and use the person's name.

Then add a Slack channel to it with one session per Slack channel, publish the channel, and give me the link to the channel settings.
```

Codex creates the agent, adds and publishes the Slack channel, and returns the link where you click **Add to Slack**.

## What the Agent Can Do in Slack

- **Its own Slack app**: each agent appears under its own name and avatar. Renaming the agent updates the app.
- **Threads, channels and direct messages**: one session per thread, per channel, or per user. Slack's agent pane is available too, with streaming replies.
- **Response policy**: answer every message, only mentions, or only messages relevant to the agent's job.
- **Slack actions**: react, update its own messages, share files in the thread, and look up users.
- **Approval in the thread**: when a tool call needs approval, Slack shows Approve and Decline buttons that only the person who asked can use.
- **Schedules**: post a morning briefing or a weekly review into a Slack conversation.
- **A record of every turn**: each conversation is an Everruns session you can open, trace and debug.

## Use Case Examples

- **Company handbook**: policies, holidays, benefits and who to ask
- **IT help desk**: account access, laptops and common fixes, with approval before any change
- **On-call helper**: runbooks and dashboards in the incident channel
- **Onboarding buddy**: answers a new hire's questions in a direct message during their first weeks
- **Household routines**: chores, screen time and a daily briefing

## What You Need

To try it quickly, use [the demo](/demo/) at app.everruns.com: it has models set up, so you can connect a Slack workspace and add an agent in a few minutes. To run Slack agents for your team, use your own Everruns:

- A running Everruns, [with Docker Compose](https://everruns.com/docs/getting-started/docker-compose/) or [on Kubernetes](https://everruns.com/docs/getting-started/kubernetes/), reachable at a public HTTPS address so Slack can deliver events to it.
- A [model provider](https://everruns.com/docs/providers/) configured in Everruns.
- A Slack workspace connected to your Everruns organization, in **Settings > Slack workspaces**.
- Permission to install apps in that workspace, or an admin who approves the request.

## Further Reading

- [Build Agents for Slack](https://everruns.com/docs/integrations/slack/), the Slack overview in the docs
- [Publish an Agent to Slack](https://everruns.com/docs/how-to/publish-to-slack/), step by step
- [Slack reference](https://everruns.com/docs/capabilities/slack/), scopes, session routing, actions and troubleshooting
- [Use Everruns in AI tools](https://everruns.com/docs/getting-started/use-in-ai-tools/), Codex, Claude Code and other MCP clients
- [Define agents as files](https://everruns.com/docs/how-to/define-agents-as-files/), the `agent.toml` format

---

## Machine-readable entry points

- [/llms.txt](https://everruns.com/llms.txt): index of this site written for language models.
- [/llms-full.txt](https://everruns.com/llms-full.txt): every page on this site as one document.
- [/auth.md](https://everruns.com/auth.md): how an agent obtains Everruns API credentials.
- [/.well-known/ai-catalog.json](https://everruns.com/.well-known/ai-catalog.json): ARD capability manifest.
- [/.well-known/mcp/server-card.json](https://everruns.com/.well-known/mcp/server-card.json): MCP server card for Everruns Managed Demo.
- [/.well-known/agent-skills/index.json](https://everruns.com/.well-known/agent-skills/index.json): agent skills for working with Everruns.

Any page on this site serves Markdown to `Accept: text/markdown`.
