Getting Started

@swa works wherever your team works. You can use it:

In Public Channels

- Simply @swa to add it to the channel and start a conversation!
- Collaborate with Swa with teammates, and access your custom-agents.
- Public channel context is available to everyone in the channel. This means @swa can reference previous questions and give smarter, more relevant answers—especially when multiple team members are involved.
- @swa can access context from any channel you and @swa are in, including accessing public channels in private and DM's (Feature Coming Soon!)

In Private Channels

- - These are limited to you and those you invite to that channel.
  - Content in private channels are 100% private and not shared across-channel.
  - Anything discussed in a private channel is limited to those in the channel and will not 'escape' the channel.

Pro Tip: Use private channels to create topics limited to just you!

Direct Messages (DMs)

- No need to @swa in DM's! Simply talk to swa as you would an assistant.
- Just like private channels, content in Direct Messages is 100% private, not shared in Public channels, and is completely separate. This context is never shared across channels or other locations and is only accessible to you.
- Each chat works the same as in Channels with a threaded conversation.

Note: Agents created are available anywhere, not just in DM's or private channels.

Create Agents

Custom agents in Swa let you design purpose-built AI assistants that think, speak, and act exactly how you need. Whether you're automating customer support, building a knowledge helper, writing content, or handling research tasks, Swa lets you pick the "brain" behind the agent—choosing from top-tier large language models—and customize behavior, tone, and scope with precision.

Example use cases:

A customer support agent using Claude for safe, policy-aligned responses
A marketing copywriter powered by ChatGPT for tone-matching and creative content
A technical research bot leveraging Perplexity for citation-backed results and depth
A conversational concierge running on Grok for lively, persistent user engagement

It's as simple as asking to make an agent and providing some basic instructions.

Step 1: Check Existing Agents

Before diving in, make sure your agent will have a unique name:

Ask swa "@swa what are my agents?"
Verify your proposed name isn't already taken
Pro tip: Consider using a themed name relevant to your agent's purpose!

Step 2: Craft Your Agent's Description

Time to give your agent its personality and purpose:

Option 1: Ask for help with "@swa Help me write a description for an agent that does [XYZ]"

Option 2:Write it yourself, making sure to include:
  - Clear purpose statement
  - Content focus areas
  - Communication style preferences
  - Any specific instructions or behaviors

Step 3: Determine the best AI model

Use @swa to find the perfect "brain" for your agent.

Available Models:

Claude from Anthropic
- Known for ethical reasoning and safety features
- Well-suited for applications requiring adherence to guidelines and sensitivity
ChatGPT from OpenAI
- Versatile and user-friendly
- Great for everyday tasks from casual conversation to problem-solving and content creation
Perplexity
- Ideal for complex, nuanced inquiries
- Excels in generating highly detailed and context-aware responses
Grok from xAI
- Focuses on conversational understanding
- Perfect for chatbots needing to maintain engaging dialogues over extended interactions

You can use Swa to help you determine which model would be best for your task.

Ask "@swa which AI model is best for [your specific task]?"
- "Which AI model is best for customer support?"
- "Which AI model is best for creating educational materials?"
For searching Live information, you'll want to use Perplexity.

Step 4: Create your agent

Now its time to create your agent.

If you've asked all of these questions in a single chat with Swa, then just ask "@swa create my agent using [Model] and name it [agent name].

Swa should already have the description since you worked on it with swa, and it will instantly create your agent!

You can ask "@swa what are my agents" to confirm it was created.

Alternatively, you can also ask

"@swa, create an agent named [my agent] using [model] and with this description [description]

Congratulations!
You've now created your custom AI agent on Swa! You've built an AI agent perfectly tailored to your needs.

Sww what @swa can do

Examples In Action

Bring powerful, purpose-driven AI into your Slack with zero setup and immediate results. Here’s a snapshot of example use cases, powered by Custom Agents and multi-model intelligence.

Optimize Costs

@swa analyzes your environment and delivers grouped, actionable cost-saving strategies—system consolidation, licensing, automation, and service improvements.

Great for: Operations teams, CFOs, budget reviewers.

Customer Confusion

When customers get lost, Swa provides an end-to-end escalation plan: from frontline responses to long-term product documentation changes. You get tactical triage steps, comms guidance, and product improvement feedback all in one output. Great for: Support directors, product managers, CX teams.

Manage Liability with AI-Crafted Policies

Swa produces legal-sounding but readable liability terms tailored to specific failure modes. This example covers exclusions, training mandates, arbitration policy, and customer limitations.Great for: Legal teams, compliance officers, operations leadership.

AI for Anything

Swa empowers every team with AI, no matter the task. Create marketing slogans, plan projects, or streamline HR—all in Slack with top models. Create images for your latest product, or even a short marketing video.

Whether it’s brainstorming, analytics, or content creation, Swa’s secure platform delivers AI for any need, instantly.

Swa Admin Panel

Everything you need to manage models, monitor usage, and control your platform.

The SWA Admin Panel gives you full visibility and control over your AI environment. Whether you're adjusting model access, reviewing usage trends, managing user permissions, or integrating with your tools, this dashboard is your operational command center.

From here you can:

Monitor token consumption and platform activity across users

Enable or disable access to specific LLMs (e.g., GPT-4 Mini, Claude Sonnet)

Manage billing plans and upgrade as usage grows

Connect to external platforms and integrations

Review reports, set limits, and adjust configuration with confidence

Tips & Best Practices

Whether you're just getting started or already creating advanced agents, these tips will help you get more out of Swa. We've designed Swa to be flexible and fast, but like any AI platform, there are a few behaviors and workarounds worth knowing to get the best experience.

Response Times

At Swa, we leverage a wide range of generative AI models - from quick text and research to more intensive audio, photo, and even video generation. Each model has its own response time and capabilities, and some tasks naturally take longer than others. That’s normal.

The advantage with Swa: you don’t have to wait. While one agent is working on a task, you can keep moving - ask another agent a different question in the same thread or start a brand‑new one!

Photo & File Uploads

Right now, Swa supports one photo or file upload at a time. File size and content can affect processing - for example, images or standard documents usually work fine, but very large spreadsheets can exceed token limits and may fail to process.

We’re actively working on improvements to handle larger and multiple files, and a solution is on the way.

Thread Length

Very long Slack threads (with dozens of messages or multiple file uploads) may sometimes fail to process because the models can’t handle that much data at once.

If this happens, simply start a new thread to continue the conversation. We’re actively developing a new context and history architecture that will not only resolve this issue but also unlock several powerful new features.

Hallucinations & Accuracy

Like all AI systems, generative models can sometimes produce inaccurate or incomplete responses - a phenomenon often called “hallucination.”

The advantage with Swa: you’re not limited to a single model. You can ask multiple agents or models the same question, compare their outputs, and even consolidate the answers. This approach significantly reduces the risk of errors and gives you a broader perspective.

Tip: For everyday tasks, multi‑model responses often provide the accuracy you need. For mission‑critical information, we recommend verifying results before acting on them.

Roadmap & Feedback

Coming Soon!

Photo & Video Creators

77%

Teams Support

58%

More Apps & Integrations

85%

We’re actively working on features designed to make Swa even more powerful and collaborative:

Photo & Video Creators - Generate images and short videos with Swa for faster content creation.
Team Collaboration & Shared Agents - Create agents your whole team can use!
More Apps & Integrations - We're connecting other GenAI tools, models, and apps to Swa! Check back often to see what we've connected!

We Want to Hear From You

Your experience helps us make Swa better every week. Whether it’s a feature request, a bug you’ve spotted, or a new use case we haven’t considered, your input goes directly into shaping our roadmap.