Agents
An agent is a deployed instance of your product's AI copilot on a specific channel. Your product defines the knowledge base, tools, and branding. Each agent inherits all of that and adds channel-specific configuration — tone, tool access, response format, and LLM settings.
A product can have multiple agents. Each one serves a different channel while sharing the same underlying knowledge and tools.
Channels
Every channel has the same capabilities. Pick whichever ones put your agent where your users and team already are.
| Channel | What it is |
|---|---|
| Web Widget | Embedded panel in your web app via the JavaScript SDK |
| Slack | Bot in a Slack workspace, responds to mentions and DMs |
| Discord | Bot in a Discord server, responds to mentions, DMs, and slash commands |
| API | Headless HTTP endpoint, no UI |
| MCP Server | Model Context Protocol server for AI IDEs |
Creating an agent
- Open your product in the Pillar dashboard
- Go to Agents
- Click Add Agent and select a channel
- Give it a name and configure channel-specific settings
What agents control
Each agent can override product-level defaults for:
- Tone — professional, friendly, neutral, concise, or formal
- Tool access — allowlist and denylist per tool, plus DMs vs public context for Slack and Discord
- Knowledge scope — all sources, internal only, external only, or a specific selection
- Response format — max tokens, source citations, follow-up suggestions
- LLM model — override the product's default model and temperature
- Guidance — channel-specific instructions appended to product-level guidance (e.g., "Keep responses under 3 sentences" for Slack)
Fields left empty on the agent fall through to the product's defaults.
User identity
To map channel users to accounts in your app, see Identity Linking.
Next steps
- Web Widget — Set up the browser SDK
- Slack — Connect a Slack bot
- Discord — Connect a Discord bot
- API — Use the headless API
- MCP Server — Expose your agent to AI IDEs