Agents – AI Teammates

AI teammates for channels, apps and users

Conclude Agents are AI teammates that can be deployed in Slack or Teams channels and added to apps. Once added, an agent joins the conversation like any other member: it can answer questions, help summarize and resolve tickets, and show the right information from your knowledge base.

Each agent is built from four things: a model (the AI that powers it), a role and instructions (what it does and how it behaves), optional knowledge (your documents, websites, and FAQs), and a deployment (the channels and apps where it is active).

With Conclude Agents you can:

  • Deploy AI teammates into Slack or Teams channels, or attach them to Conclude apps to help handle tickets
  • Choose the model that powers each agent (e.g., Claude Opus 5, Claude Sonnet 5, or other providers) – optionally using your own provider API keys
  • Give each agent a role and detailed instructions so it responds in your tone and stays on task
  • Ground agents in your own knowledge — upload documents or add website and FAQ URLs so answers are accurate and cited
  • Control when an agent speaks — only when @mentioned, or proactively when it can help
  • Set guardrails such as the maximum number of comments per ticket and the maximum response length

To use Conclude Agents, you will need:

  • An active Conclude account — see our Quickstart Guides to get started
  • Agents enabled under Admin console > Solutions
  • Optional: Apps enabled if you want agents to work with Conclude Apps for tickets and issues
  • Optional: Connect enabled if you want agents to work in Slack channels connected to Microsoft Teams
  • Optional: provide your own model provider API key under Admin console > Agents> All agents
  • Optional: one or more Slack channels or Conclude apps where you want the agent to work

Start here

In this guide, we cover:

  • How Conclude Agents work
  • Adding a provider secret (optional)
  • Creating a knowledge collection (optional)
  • Creating an agent – manually or with AI
  • Configuring the agent – role, model, mode, instructions, knowledge, and limits
  • Deploying the agent to channels and apps
  • Testing your agent

How Conclude Agents work

An agent is a configurable AI teammate. You give it a model, a role, and instructions; optionally connect it to one or more knowledge collections; then deploy it where your team works

Agents can be deployed in two places:

  • Channels — the agent is added to a Slack or Teams channel and takes part in the conversation there
  • Apps — the agent is attached to a Conclude app and answers comments on tickets when it is named in one

Agents respond in one of two ways:

  • Mention — the agent joins a conversation only when @mentioned (e.g. @Ada)
  • Autonomous — the agent joins a conversation unprompted, speaking up on its own when it can help

💡 You can connect an agent to a dedicated user account, so its messages appear under a recognizable name and avatar (for example, “Ada”) rather than a generic bot.

Step 1: Choose how agents are powered (Provider secrets)

Every agent runs on a model from one of the providers we currently support: Anthropic and OpenAI. You can use the included agent usage or connect your own API key. Adding your own key is optional.

Use the included agent usage

No setup is required. Agents can run using Conclude’s included usage.

  • Go to Admin console > Agents > All agents to see how much of your included agent usage has been used this month.

     

agents overview where ai teammates and their models are managed

 

Use your own API key

If you prefer to use your own provider account:

  • Go to Admin console > Agents > All agents

  • Click Add key

  • Choose Anthropic or OpenAI

  • Enter your API key, and save

     

switching agent usage from conclude keys to your own provider keys

 

Agents using that provider will then run on your API key, and usage will be billed directly by the provider.

Step 2: Create a knowledge collection (optional)

Knowledge collections let agents answer from your own material – product docs, your website, or FAQs – instead of general knowledge alone.

  • Go to Admin console > Agents > Knowledge and click Add collection

 

knowledge collections that ai teammates can draw their answers from

 

  • Give the knowledge collection a name and description

 

adding files or urls as sources to a new knowledge collection

 

  • Add sources: upload files (Markdown, text, HTML, DOCX — 10 MB max) or add a URL with a title
  • Each source shows a status of Ready once it has been processed

 

a knowledge collection with its sources indexed and ready to use

 

💡 You can mix file and URL sources in one collection e.g., a knowledge base document plus your website and FAQ pages. You will connect the collection to an agent in Step 4.

Step 3: Create an agent

Go to Admin console > Agents > All agents. Existing agents are listed with their role, model, and where they are deployed.

 

included agent usage for the month before adding your own keys

 

There are two ways to create a new agent:

Option 1: Create with AI

  • Click Create with AI
  • Describe the kind of agent you want (e.g., “A friendly support agent for a SaaS product that helps users troubleshoot issues”)
  • Optionally add a Company name and Tone
  • Click Generate – Conclude drafts the role and instructions for you, which you can refine

 

describing an ai teammate in plain words to have it generated

 

Option 2: Add an agent manually

  • Click Add agent
  • Enter a Name and a Role
  • Choose a Model – the note below the dropdown shows which provider key it uses
  • Write the agent’s Instructions, then click Create

 

setting up an ai teammate name role model and instructions by hand

 

Step 4: Configure the agent

Open an agent from the list to fine-tune it across five tabs: General, Deployment, Instructions, Knowledge, and Advanced.

General

  • Role – a short description of what the agent does
  • Model – the model powering the agent (e.g., Claude Opus 5)
  • Mode – how the agent decides when to speak, with two options:
  • Mention – joins a conversation only when @mentioned in it
  • Autonomous – joins a conversation unprompted
  • Off — never listens or respond
  • Identity— optional. Connect a dedicated account so the agent appears as a channel member, uses that account’s name and profile picture, and can be found through @ autocomplete.
    Without a connected account, the agent keeps its own name and avatar and can still respond when users mention its name.

 

choosing the model and whether an ai teammate replies only when mentioned

How to connect an agent to a dedicated user account

  • On the agent’s General tab, click Linked agent Connect to user account

  • Under User account, select the account created for this agent, then click Link

     

giving an ai teammate its own account so it can join channels

⚠ Use a dedicated account made only for this agent. When connected, the agent posts as that account, is @mentioned as it, and takes its name and picture, so a personal account would sign everything the agent says with that person’s name.

Why connect a user account:

  • The agent appears as a channel member and is reachable through @ auto-completion, like any teammate
  • Its messages appear under a recognizable name and avatar instead of a generic bot identity
  • Without a connected account, the agent still responds in channels, but people must @mention it by its exact name (for example, @Morgan) — it does not appear in @ auto-completion and is not listed as a channel member
  • When an agent is connected to a user account, it is automatically present in every Slack or Teams channel that account is a member of. You can control how the agent behaves in each channel from the Deployment tab.

💡 Disconnecting reverts the agent to its own name and an avatar drawn from its initials, which reads as an agent rather than a person. It stops appearing in @ auto-completion and the channel member list, but it still responds when @mentioned by name

Instructions

This is the agent’s system prompt — it defines behavior, tone, and boundaries. Strong instructions usually cover:

  • The agent’s job and core behaviors
  • When to speak up and when to stay quiet
  • Tone and formatting (e.g., short answers, friendly phrasing, bulleted steps)
  • Boundaries – for example, don’t invent features, and escalate when unsure
  • Citation style, if the agent draws on knowledge sources

 

writing the instructions that shape how an ai teammate answers

 

Knowledge

  • Select the knowledge collections this agent should have access to (e.g., your company’s support docs, legal agreements)
  • The agent uses these collections to ground its answers

 

choosing which knowledge collections an ai teammate may use

 

Advanced

  • Max comments – the most comments the agent will post on a single ticket, no matter how much activity it sees (default 20 comments)
  • Max tokens – the maximum length of each response (default 1024 characters)

 

limiting how many comments and tokens an ai teammate spends per ticket

 

💡 Remember to click Save after making changes on any tab.

Step 5: Deploy the agent

Open the agent and go to the Deployment tab. Here you choose where the agent is present and control how it behaves in each channel or app.

 

where an ai teammate is deployed across slack teams channels and apps

 

Adding an agent to a channel

If the agent is connected to a dedicated user account, it automatically appears in every channel where that account is already a member. You can control how the agent behaves in each of these channels from the Deployment tab.

To add the agent to another channel:

  • Click Add to channel
  • Choose the Slack workspace and channel

 

adding an ai teammate to a slack channel so it joins the conversation

Choose how the agent behaves in each channel:

  • Agent default — follows the mode configured for the agent
  • Mention — responds only when @mentioned
  • Autonomous — can respond proactively without being mentioned
  • Off — does not listen or respond in that channel

 

where an ai teammate is deployed across slack teams channels and apps

Adding an agent to an app

Agents can also work inside a Conclude app to help with tickets. There are two entry points.

  • From the agent – on the agent’s Deployment tab, click Add to app and select an app

 

adding an ai teammate to an app so it answers ticket comments

 

  • From the app – go to Apps in the Dashboard, click the Edit (pencil) icon next to the app, open the Agents tab, and click Add agent

 

which ai teammates an app uses and what each one responds to

 

💡 Agents are defined in the admin console; in the app editor you choose which of them respond in this app, and how.

Configure how the agent responds in the app

When you add an agent to an app, you can set the following:

  • Interaction mode — Default (use whatever the agent itself is set to), Mention (only when named in a comment), or Autonomous (unprompted)

 

choosing whether an ai teammate waits to be mentioned or acts on its own

 

  • The events the agent responds to: New ticket, Comment, Edit, and Conclude

 

picking which ticket events an ai teammate responds to

Step 6: Test your agent

  • Go to a channel or ticket where the agent is deployed
  • @mention the agent (or post a relevant message if it is set to respond proactively)
  • Confirm it replies in the right tone, pulls from the correct knowledge, and respects your limits

💡 If the agent doesn’t respond, check that the provider secret is valid, the agent is deployed to that channel or app, and – for apps – that it has been named on the ticket and the right events are enabled in the app editor.

 

Connect. Collaborate. Conclude